You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hbase.apache.org by gi...@apache.org on 2018/04/19 14:46:38 UTC
[11/43] hbase-site git commit: Published site at
556b22374423ff087c0583d02ae4298d4d4f2e6b.
http://git-wip-us.apache.org/repos/asf/hbase-site/blob/ede30993/devapidocs/src-html/org/apache/hadoop/hbase/util/HBaseFsck.RegionBoundariesInformation.html
----------------------------------------------------------------------
diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/util/HBaseFsck.RegionBoundariesInformation.html b/devapidocs/src-html/org/apache/hadoop/hbase/util/HBaseFsck.RegionBoundariesInformation.html
index c370eb9..e1bc325 100644
--- a/devapidocs/src-html/org/apache/hadoop/hbase/util/HBaseFsck.RegionBoundariesInformation.html
+++ b/devapidocs/src-html/org/apache/hadoop/hbase/util/HBaseFsck.RegionBoundariesInformation.html
@@ -6,7 +6,7 @@
</head>
<body>
<div class="sourceContainer">
-<pre><span class="sourceLineNo">001</span>/**<a name="line.1"></a>
+<pre><span class="sourceLineNo">001</span>/*<a name="line.1"></a>
<span class="sourceLineNo">002</span> * Licensed to the Apache Software Foundation (ASF) under one<a name="line.2"></a>
<span class="sourceLineNo">003</span> * or more contributor license agreements. See the NOTICE file<a name="line.3"></a>
<span class="sourceLineNo">004</span> * distributed with this work for additional information<a name="line.4"></a>
@@ -144,5002 +144,5047 @@
<span class="sourceLineNo">136</span>import org.apache.hadoop.util.ReflectionUtils;<a name="line.136"></a>
<span class="sourceLineNo">137</span>import org.apache.hadoop.util.Tool;<a name="line.137"></a>
<span class="sourceLineNo">138</span>import org.apache.hadoop.util.ToolRunner;<a name="line.138"></a>
-<span class="sourceLineNo">139</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.139"></a>
-<span class="sourceLineNo">140</span>import org.apache.yetus.audience.InterfaceStability;<a name="line.140"></a>
-<span class="sourceLineNo">141</span>import org.apache.zookeeper.KeeperException;<a name="line.141"></a>
-<span class="sourceLineNo">142</span>import org.slf4j.Logger;<a name="line.142"></a>
-<span class="sourceLineNo">143</span>import org.slf4j.LoggerFactory;<a name="line.143"></a>
-<span class="sourceLineNo">144</span><a name="line.144"></a>
-<span class="sourceLineNo">145</span>import org.apache.hbase.thirdparty.com.google.common.base.Joiner;<a name="line.145"></a>
-<span class="sourceLineNo">146</span>import org.apache.hbase.thirdparty.com.google.common.base.Preconditions;<a name="line.146"></a>
-<span class="sourceLineNo">147</span>import org.apache.hbase.thirdparty.com.google.common.collect.ImmutableList;<a name="line.147"></a>
-<span class="sourceLineNo">148</span>import org.apache.hbase.thirdparty.com.google.common.collect.Lists;<a name="line.148"></a>
-<span class="sourceLineNo">149</span>import org.apache.hbase.thirdparty.com.google.common.collect.Multimap;<a name="line.149"></a>
-<span class="sourceLineNo">150</span>import org.apache.hbase.thirdparty.com.google.common.collect.Ordering;<a name="line.150"></a>
-<span class="sourceLineNo">151</span>import org.apache.hbase.thirdparty.com.google.common.collect.TreeMultimap;<a name="line.151"></a>
-<span class="sourceLineNo">152</span><a name="line.152"></a>
-<span class="sourceLineNo">153</span>import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;<a name="line.153"></a>
-<span class="sourceLineNo">154</span>import org.apache.hadoop.hbase.shaded.protobuf.generated.AdminProtos.AdminService.BlockingInterface;<a name="line.154"></a>
-<span class="sourceLineNo">155</span><a name="line.155"></a>
-<span class="sourceLineNo">156</span>/**<a name="line.156"></a>
-<span class="sourceLineNo">157</span> * HBaseFsck (hbck) is a tool for checking and repairing region consistency and<a name="line.157"></a>
-<span class="sourceLineNo">158</span> * table integrity problems in a corrupted HBase.<a name="line.158"></a>
-<span class="sourceLineNo">159</span> * <p><a name="line.159"></a>
-<span class="sourceLineNo">160</span> * Region consistency checks verify that hbase:meta, region deployment on region<a name="line.160"></a>
-<span class="sourceLineNo">161</span> * servers and the state of data in HDFS (.regioninfo files) all are in<a name="line.161"></a>
-<span class="sourceLineNo">162</span> * accordance.<a name="line.162"></a>
-<span class="sourceLineNo">163</span> * <p><a name="line.163"></a>
-<span class="sourceLineNo">164</span> * Table integrity checks verify that all possible row keys resolve to exactly<a name="line.164"></a>
-<span class="sourceLineNo">165</span> * one region of a table. This means there are no individual degenerate<a name="line.165"></a>
-<span class="sourceLineNo">166</span> * or backwards regions; no holes between regions; and that there are no<a name="line.166"></a>
-<span class="sourceLineNo">167</span> * overlapping regions.<a name="line.167"></a>
-<span class="sourceLineNo">168</span> * <p><a name="line.168"></a>
-<span class="sourceLineNo">169</span> * The general repair strategy works in two phases:<a name="line.169"></a>
-<span class="sourceLineNo">170</span> * <ol><a name="line.170"></a>
-<span class="sourceLineNo">171</span> * <li> Repair Table Integrity on HDFS. (merge or fabricate regions)<a name="line.171"></a>
-<span class="sourceLineNo">172</span> * <li> Repair Region Consistency with hbase:meta and assignments<a name="line.172"></a>
-<span class="sourceLineNo">173</span> * </ol><a name="line.173"></a>
-<span class="sourceLineNo">174</span> * <p><a name="line.174"></a>
-<span class="sourceLineNo">175</span> * For table integrity repairs, the tables' region directories are scanned<a name="line.175"></a>
-<span class="sourceLineNo">176</span> * for .regioninfo files. Each table's integrity is then verified. If there<a name="line.176"></a>
-<span class="sourceLineNo">177</span> * are any orphan regions (regions with no .regioninfo files) or holes, new<a name="line.177"></a>
-<span class="sourceLineNo">178</span> * regions are fabricated. Backwards regions are sidelined as well as empty<a name="line.178"></a>
-<span class="sourceLineNo">179</span> * degenerate (endkey==startkey) regions. If there are any overlapping regions,<a name="line.179"></a>
-<span class="sourceLineNo">180</span> * a new region is created and all data is merged into the new region.<a name="line.180"></a>
-<span class="sourceLineNo">181</span> * <p><a name="line.181"></a>
-<span class="sourceLineNo">182</span> * Table integrity repairs deal solely with HDFS and could potentially be done<a name="line.182"></a>
-<span class="sourceLineNo">183</span> * offline -- the hbase region servers or master do not need to be running.<a name="line.183"></a>
-<span class="sourceLineNo">184</span> * This phase can eventually be used to completely reconstruct the hbase:meta table in<a name="line.184"></a>
-<span class="sourceLineNo">185</span> * an offline fashion.<a name="line.185"></a>
-<span class="sourceLineNo">186</span> * <p><a name="line.186"></a>
-<span class="sourceLineNo">187</span> * Region consistency requires three conditions -- 1) valid .regioninfo file<a name="line.187"></a>
-<span class="sourceLineNo">188</span> * present in an HDFS region dir, 2) valid row with .regioninfo data in META,<a name="line.188"></a>
-<span class="sourceLineNo">189</span> * and 3) a region is deployed only at the regionserver that was assigned to<a name="line.189"></a>
-<span class="sourceLineNo">190</span> * with proper state in the master.<a name="line.190"></a>
-<span class="sourceLineNo">191</span> * <p><a name="line.191"></a>
-<span class="sourceLineNo">192</span> * Region consistency repairs require hbase to be online so that hbck can<a name="line.192"></a>
-<span class="sourceLineNo">193</span> * contact the HBase master and region servers. The hbck#connect() method must<a name="line.193"></a>
-<span class="sourceLineNo">194</span> * first be called successfully. Much of the region consistency information<a name="line.194"></a>
-<span class="sourceLineNo">195</span> * is transient and less risky to repair.<a name="line.195"></a>
-<span class="sourceLineNo">196</span> * <p><a name="line.196"></a>
-<span class="sourceLineNo">197</span> * If hbck is run from the command line, there are a handful of arguments that<a name="line.197"></a>
-<span class="sourceLineNo">198</span> * can be used to limit the kinds of repairs hbck will do. See the code in<a name="line.198"></a>
-<span class="sourceLineNo">199</span> * {@link #printUsageAndExit()} for more details.<a name="line.199"></a>
-<span class="sourceLineNo">200</span> */<a name="line.200"></a>
-<span class="sourceLineNo">201</span>@InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.TOOLS)<a name="line.201"></a>
-<span class="sourceLineNo">202</span>@InterfaceStability.Evolving<a name="line.202"></a>
-<span class="sourceLineNo">203</span>public class HBaseFsck extends Configured implements Closeable {<a name="line.203"></a>
-<span class="sourceLineNo">204</span> public static final long DEFAULT_TIME_LAG = 60000; // default value of 1 minute<a name="line.204"></a>
-<span class="sourceLineNo">205</span> public static final long DEFAULT_SLEEP_BEFORE_RERUN = 10000;<a name="line.205"></a>
-<span class="sourceLineNo">206</span> private static final int MAX_NUM_THREADS = 50; // #threads to contact regions<a name="line.206"></a>
-<span class="sourceLineNo">207</span> private static boolean rsSupportsOffline = true;<a name="line.207"></a>
-<span class="sourceLineNo">208</span> private static final int DEFAULT_OVERLAPS_TO_SIDELINE = 2;<a name="line.208"></a>
-<span class="sourceLineNo">209</span> private static final int DEFAULT_MAX_MERGE = 5;<a name="line.209"></a>
-<span class="sourceLineNo">210</span> private static final String TO_BE_LOADED = "to_be_loaded";<a name="line.210"></a>
-<span class="sourceLineNo">211</span> private static final String HBCK_LOCK_FILE = "hbase-hbck.lock";<a name="line.211"></a>
-<span class="sourceLineNo">212</span> private static final int DEFAULT_MAX_LOCK_FILE_ATTEMPTS = 5;<a name="line.212"></a>
-<span class="sourceLineNo">213</span> private static final int DEFAULT_LOCK_FILE_ATTEMPT_SLEEP_INTERVAL = 200; // milliseconds<a name="line.213"></a>
-<span class="sourceLineNo">214</span> private static final int DEFAULT_LOCK_FILE_ATTEMPT_MAX_SLEEP_TIME = 5000; // milliseconds<a name="line.214"></a>
-<span class="sourceLineNo">215</span> // We have to set the timeout value > HdfsConstants.LEASE_SOFTLIMIT_PERIOD.<a name="line.215"></a>
-<span class="sourceLineNo">216</span> // In HADOOP-2.6 and later, the Namenode proxy now created with custom RetryPolicy for<a name="line.216"></a>
-<span class="sourceLineNo">217</span> // AlreadyBeingCreatedException which is implies timeout on this operations up to<a name="line.217"></a>
-<span class="sourceLineNo">218</span> // HdfsConstants.LEASE_SOFTLIMIT_PERIOD (60 seconds).<a name="line.218"></a>
-<span class="sourceLineNo">219</span> private static final int DEFAULT_WAIT_FOR_LOCK_TIMEOUT = 80; // seconds<a name="line.219"></a>
-<span class="sourceLineNo">220</span> private static final int DEFAULT_MAX_CREATE_ZNODE_ATTEMPTS = 5;<a name="line.220"></a>
-<span class="sourceLineNo">221</span> private static final int DEFAULT_CREATE_ZNODE_ATTEMPT_SLEEP_INTERVAL = 200; // milliseconds<a name="line.221"></a>
-<span class="sourceLineNo">222</span> private static final int DEFAULT_CREATE_ZNODE_ATTEMPT_MAX_SLEEP_TIME = 5000; // milliseconds<a name="line.222"></a>
-<span class="sourceLineNo">223</span><a name="line.223"></a>
-<span class="sourceLineNo">224</span> /**********************<a name="line.224"></a>
-<span class="sourceLineNo">225</span> * Internal resources<a name="line.225"></a>
-<span class="sourceLineNo">226</span> **********************/<a name="line.226"></a>
-<span class="sourceLineNo">227</span> private static final Logger LOG = LoggerFactory.getLogger(HBaseFsck.class.getName());<a name="line.227"></a>
-<span class="sourceLineNo">228</span> private ClusterMetrics status;<a name="line.228"></a>
-<span class="sourceLineNo">229</span> private ClusterConnection connection;<a name="line.229"></a>
-<span class="sourceLineNo">230</span> private Admin admin;<a name="line.230"></a>
-<span class="sourceLineNo">231</span> private Table meta;<a name="line.231"></a>
-<span class="sourceLineNo">232</span> // threads to do ||izable tasks: retrieve data from regionservers, handle overlapping regions<a name="line.232"></a>
-<span class="sourceLineNo">233</span> protected ExecutorService executor;<a name="line.233"></a>
-<span class="sourceLineNo">234</span> private long startMillis = EnvironmentEdgeManager.currentTime();<a name="line.234"></a>
-<span class="sourceLineNo">235</span> private HFileCorruptionChecker hfcc;<a name="line.235"></a>
-<span class="sourceLineNo">236</span> private int retcode = 0;<a name="line.236"></a>
-<span class="sourceLineNo">237</span> private Path HBCK_LOCK_PATH;<a name="line.237"></a>
-<span class="sourceLineNo">238</span> private FSDataOutputStream hbckOutFd;<a name="line.238"></a>
-<span class="sourceLineNo">239</span> // This lock is to prevent cleanup of balancer resources twice between<a name="line.239"></a>
-<span class="sourceLineNo">240</span> // ShutdownHook and the main code. We cleanup only if the connect() is<a name="line.240"></a>
-<span class="sourceLineNo">241</span> // successful<a name="line.241"></a>
-<span class="sourceLineNo">242</span> private final AtomicBoolean hbckLockCleanup = new AtomicBoolean(false);<a name="line.242"></a>
-<span class="sourceLineNo">243</span><a name="line.243"></a>
-<span class="sourceLineNo">244</span> /***********<a name="line.244"></a>
-<span class="sourceLineNo">245</span> * Options<a name="line.245"></a>
-<span class="sourceLineNo">246</span> ***********/<a name="line.246"></a>
-<span class="sourceLineNo">247</span> private static boolean details = false; // do we display the full report<a name="line.247"></a>
-<span class="sourceLineNo">248</span> private long timelag = DEFAULT_TIME_LAG; // tables whose modtime is older<a name="line.248"></a>
-<span class="sourceLineNo">249</span> private static boolean forceExclusive = false; // only this hbck can modify HBase<a name="line.249"></a>
-<span class="sourceLineNo">250</span> private boolean fixAssignments = false; // fix assignment errors?<a name="line.250"></a>
-<span class="sourceLineNo">251</span> private boolean fixMeta = false; // fix meta errors?<a name="line.251"></a>
-<span class="sourceLineNo">252</span> private boolean checkHdfs = true; // load and check fs consistency?<a name="line.252"></a>
-<span class="sourceLineNo">253</span> private boolean fixHdfsHoles = false; // fix fs holes?<a name="line.253"></a>
-<span class="sourceLineNo">254</span> private boolean fixHdfsOverlaps = false; // fix fs overlaps (risky)<a name="line.254"></a>
-<span class="sourceLineNo">255</span> private boolean fixHdfsOrphans = false; // fix fs holes (missing .regioninfo)<a name="line.255"></a>
-<span class="sourceLineNo">256</span> private boolean fixTableOrphans = false; // fix fs holes (missing .tableinfo)<a name="line.256"></a>
-<span class="sourceLineNo">257</span> private boolean fixVersionFile = false; // fix missing hbase.version file in hdfs<a name="line.257"></a>
-<span class="sourceLineNo">258</span> private boolean fixSplitParents = false; // fix lingering split parents<a name="line.258"></a>
-<span class="sourceLineNo">259</span> private boolean removeParents = false; // remove split parents<a name="line.259"></a>
-<span class="sourceLineNo">260</span> private boolean fixReferenceFiles = false; // fix lingering reference store file<a name="line.260"></a>
-<span class="sourceLineNo">261</span> private boolean fixHFileLinks = false; // fix lingering HFileLinks<a name="line.261"></a>
-<span class="sourceLineNo">262</span> private boolean fixEmptyMetaCells = false; // fix (remove) empty REGIONINFO_QUALIFIER rows<a name="line.262"></a>
-<span class="sourceLineNo">263</span> private boolean fixReplication = false; // fix undeleted replication queues for removed peer<a name="line.263"></a>
-<span class="sourceLineNo">264</span> private boolean fixAny = false; // Set to true if any of the fix is required.<a name="line.264"></a>
-<span class="sourceLineNo">265</span><a name="line.265"></a>
-<span class="sourceLineNo">266</span> // limit checking/fixes to listed tables, if empty attempt to check/fix all<a name="line.266"></a>
-<span class="sourceLineNo">267</span> // hbase:meta are always checked<a name="line.267"></a>
-<span class="sourceLineNo">268</span> private Set<TableName> tablesIncluded = new HashSet<>();<a name="line.268"></a>
-<span class="sourceLineNo">269</span> private int maxMerge = DEFAULT_MAX_MERGE; // maximum number of overlapping regions to merge<a name="line.269"></a>
-<span class="sourceLineNo">270</span> // maximum number of overlapping regions to sideline<a name="line.270"></a>
-<span class="sourceLineNo">271</span> private int maxOverlapsToSideline = DEFAULT_OVERLAPS_TO_SIDELINE;<a name="line.271"></a>
-<span class="sourceLineNo">272</span> private boolean sidelineBigOverlaps = false; // sideline overlaps with >maxMerge regions<a name="line.272"></a>
-<span class="sourceLineNo">273</span> private Path sidelineDir = null;<a name="line.273"></a>
-<span class="sourceLineNo">274</span><a name="line.274"></a>
-<span class="sourceLineNo">275</span> private boolean rerun = false; // if we tried to fix something, rerun hbck<a name="line.275"></a>
-<span class="sourceLineNo">276</span> private static boolean summary = false; // if we want to print less output<a name="line.276"></a>
-<span class="sourceLineNo">277</span> private boolean checkMetaOnly = false;<a name="line.277"></a>
-<span class="sourceLineNo">278</span> private boolean checkRegionBoundaries = false;<a name="line.278"></a>
-<span class="sourceLineNo">279</span> private boolean ignorePreCheckPermission = false; // if pre-check permission<a name="line.279"></a>
-<span class="sourceLineNo">280</span><a name="line.280"></a>
-<span class="sourceLineNo">281</span> /*********<a name="line.281"></a>
-<span class="sourceLineNo">282</span> * State<a name="line.282"></a>
-<span class="sourceLineNo">283</span> *********/<a name="line.283"></a>
-<span class="sourceLineNo">284</span> final private ErrorReporter errors;<a name="line.284"></a>
-<span class="sourceLineNo">285</span> int fixes = 0;<a name="line.285"></a>
-<span class="sourceLineNo">286</span><a name="line.286"></a>
-<span class="sourceLineNo">287</span> /**<a name="line.287"></a>
-<span class="sourceLineNo">288</span> * This map contains the state of all hbck items. It maps from encoded region<a name="line.288"></a>
-<span class="sourceLineNo">289</span> * name to HbckInfo structure. The information contained in HbckInfo is used<a name="line.289"></a>
-<span class="sourceLineNo">290</span> * to detect and correct consistency (hdfs/meta/deployment) problems.<a name="line.290"></a>
-<span class="sourceLineNo">291</span> */<a name="line.291"></a>
-<span class="sourceLineNo">292</span> private TreeMap<String, HbckInfo> regionInfoMap = new TreeMap<>();<a name="line.292"></a>
-<span class="sourceLineNo">293</span> // Empty regioninfo qualifiers in hbase:meta<a name="line.293"></a>
-<span class="sourceLineNo">294</span> private Set<Result> emptyRegionInfoQualifiers = new HashSet<>();<a name="line.294"></a>
-<span class="sourceLineNo">295</span><a name="line.295"></a>
-<span class="sourceLineNo">296</span> /**<a name="line.296"></a>
-<span class="sourceLineNo">297</span> * This map from Tablename -> TableInfo contains the structures necessary to<a name="line.297"></a>
-<span class="sourceLineNo">298</span> * detect table consistency problems (holes, dupes, overlaps). It is sorted<a name="line.298"></a>
-<span class="sourceLineNo">299</span> * to prevent dupes.<a name="line.299"></a>
-<span class="sourceLineNo">300</span> *<a name="line.300"></a>
-<span class="sourceLineNo">301</span> * If tablesIncluded is empty, this map contains all tables.<a name="line.301"></a>
-<span class="sourceLineNo">302</span> * Otherwise, it contains only meta tables and tables in tablesIncluded,<a name="line.302"></a>
-<span class="sourceLineNo">303</span> * unless checkMetaOnly is specified, in which case, it contains only<a name="line.303"></a>
-<span class="sourceLineNo">304</span> * the meta table<a name="line.304"></a>
-<span class="sourceLineNo">305</span> */<a name="line.305"></a>
-<span class="sourceLineNo">306</span> private SortedMap<TableName, TableInfo> tablesInfo = new ConcurrentSkipListMap<>();<a name="line.306"></a>
-<span class="sourceLineNo">307</span><a name="line.307"></a>
-<span class="sourceLineNo">308</span> /**<a name="line.308"></a>
-<span class="sourceLineNo">309</span> * When initially looking at HDFS, we attempt to find any orphaned data.<a name="line.309"></a>
-<span class="sourceLineNo">310</span> */<a name="line.310"></a>
-<span class="sourceLineNo">311</span> private List<HbckInfo> orphanHdfsDirs = Collections.synchronizedList(new ArrayList<HbckInfo>());<a name="line.311"></a>
-<span class="sourceLineNo">312</span><a name="line.312"></a>
-<span class="sourceLineNo">313</span> private Map<TableName, Set<String>> orphanTableDirs = new HashMap<>();<a name="line.313"></a>
-<span class="sourceLineNo">314</span> private Map<TableName, TableState> tableStates = new HashMap<>();<a name="line.314"></a>
-<span class="sourceLineNo">315</span> private final RetryCounterFactory lockFileRetryCounterFactory;<a name="line.315"></a>
-<span class="sourceLineNo">316</span> private final RetryCounterFactory createZNodeRetryCounterFactory;<a name="line.316"></a>
-<span class="sourceLineNo">317</span><a name="line.317"></a>
-<span class="sourceLineNo">318</span> private Map<TableName, Set<String>> skippedRegions = new HashMap<>();<a name="line.318"></a>
+<span class="sourceLineNo">139</span>import org.apache.hbase.thirdparty.com.google.common.collect.Sets;<a name="line.139"></a>
+<span class="sourceLineNo">140</span>import org.apache.yetus.audience.InterfaceAudience;<a name="line.140"></a>
+<span class="sourceLineNo">141</span>import org.apache.yetus.audience.InterfaceStability;<a name="line.141"></a>
+<span class="sourceLineNo">142</span>import org.apache.zookeeper.KeeperException;<a name="line.142"></a>
+<span class="sourceLineNo">143</span>import org.slf4j.Logger;<a name="line.143"></a>
+<span class="sourceLineNo">144</span>import org.slf4j.LoggerFactory;<a name="line.144"></a>
+<span class="sourceLineNo">145</span><a name="line.145"></a>
+<span class="sourceLineNo">146</span>import org.apache.hbase.thirdparty.com.google.common.base.Joiner;<a name="line.146"></a>
+<span class="sourceLineNo">147</span>import org.apache.hbase.thirdparty.com.google.common.base.Preconditions;<a name="line.147"></a>
+<span class="sourceLineNo">148</span>import org.apache.hbase.thirdparty.com.google.common.collect.ImmutableList;<a name="line.148"></a>
+<span class="sourceLineNo">149</span>import org.apache.hbase.thirdparty.com.google.common.collect.Lists;<a name="line.149"></a>
+<span class="sourceLineNo">150</span>import org.apache.hbase.thirdparty.com.google.common.collect.Multimap;<a name="line.150"></a>
+<span class="sourceLineNo">151</span>import org.apache.hbase.thirdparty.com.google.common.collect.Ordering;<a name="line.151"></a>
+<span class="sourceLineNo">152</span>import org.apache.hbase.thirdparty.com.google.common.collect.TreeMultimap;<a name="line.152"></a>
+<span class="sourceLineNo">153</span><a name="line.153"></a>
+<span class="sourceLineNo">154</span>import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;<a name="line.154"></a>
+<span class="sourceLineNo">155</span>import org.apache.hadoop.hbase.shaded.protobuf.generated.AdminProtos.AdminService.BlockingInterface;<a name="line.155"></a>
+<span class="sourceLineNo">156</span><a name="line.156"></a>
+<span class="sourceLineNo">157</span>/**<a name="line.157"></a>
+<span class="sourceLineNo">158</span> * HBaseFsck (hbck) is a tool for checking and repairing region consistency and<a name="line.158"></a>
+<span class="sourceLineNo">159</span> * table integrity problems in a corrupted HBase.<a name="line.159"></a>
+<span class="sourceLineNo">160</span> * <p><a name="line.160"></a>
+<span class="sourceLineNo">161</span> * Region consistency checks verify that hbase:meta, region deployment on region<a name="line.161"></a>
+<span class="sourceLineNo">162</span> * servers and the state of data in HDFS (.regioninfo files) all are in<a name="line.162"></a>
+<span class="sourceLineNo">163</span> * accordance.<a name="line.163"></a>
+<span class="sourceLineNo">164</span> * <p><a name="line.164"></a>
+<span class="sourceLineNo">165</span> * Table integrity checks verify that all possible row keys resolve to exactly<a name="line.165"></a>
+<span class="sourceLineNo">166</span> * one region of a table. This means there are no individual degenerate<a name="line.166"></a>
+<span class="sourceLineNo">167</span> * or backwards regions; no holes between regions; and that there are no<a name="line.167"></a>
+<span class="sourceLineNo">168</span> * overlapping regions.<a name="line.168"></a>
+<span class="sourceLineNo">169</span> * <p><a name="line.169"></a>
+<span class="sourceLineNo">170</span> * The general repair strategy works in two phases:<a name="line.170"></a>
+<span class="sourceLineNo">171</span> * <ol><a name="line.171"></a>
+<span class="sourceLineNo">172</span> * <li> Repair Table Integrity on HDFS. (merge or fabricate regions)<a name="line.172"></a>
+<span class="sourceLineNo">173</span> * <li> Repair Region Consistency with hbase:meta and assignments<a name="line.173"></a>
+<span class="sourceLineNo">174</span> * </ol><a name="line.174"></a>
+<span class="sourceLineNo">175</span> * <p><a name="line.175"></a>
+<span class="sourceLineNo">176</span> * For table integrity repairs, the tables' region directories are scanned<a name="line.176"></a>
+<span class="sourceLineNo">177</span> * for .regioninfo files. Each table's integrity is then verified. If there<a name="line.177"></a>
+<span class="sourceLineNo">178</span> * are any orphan regions (regions with no .regioninfo files) or holes, new<a name="line.178"></a>
+<span class="sourceLineNo">179</span> * regions are fabricated. Backwards regions are sidelined as well as empty<a name="line.179"></a>
+<span class="sourceLineNo">180</span> * degenerate (endkey==startkey) regions. If there are any overlapping regions,<a name="line.180"></a>
+<span class="sourceLineNo">181</span> * a new region is created and all data is merged into the new region.<a name="line.181"></a>
+<span class="sourceLineNo">182</span> * <p><a name="line.182"></a>
+<span class="sourceLineNo">183</span> * Table integrity repairs deal solely with HDFS and could potentially be done<a name="line.183"></a>
+<span class="sourceLineNo">184</span> * offline -- the hbase region servers or master do not need to be running.<a name="line.184"></a>
+<span class="sourceLineNo">185</span> * This phase can eventually be used to completely reconstruct the hbase:meta table in<a name="line.185"></a>
+<span class="sourceLineNo">186</span> * an offline fashion.<a name="line.186"></a>
+<span class="sourceLineNo">187</span> * <p><a name="line.187"></a>
+<span class="sourceLineNo">188</span> * Region consistency requires three conditions -- 1) valid .regioninfo file<a name="line.188"></a>
+<span class="sourceLineNo">189</span> * present in an HDFS region dir, 2) valid row with .regioninfo data in META,<a name="line.189"></a>
+<span class="sourceLineNo">190</span> * and 3) a region is deployed only at the regionserver that was assigned to<a name="line.190"></a>
+<span class="sourceLineNo">191</span> * with proper state in the master.<a name="line.191"></a>
+<span class="sourceLineNo">192</span> * <p><a name="line.192"></a>
+<span class="sourceLineNo">193</span> * Region consistency repairs require hbase to be online so that hbck can<a name="line.193"></a>
+<span class="sourceLineNo">194</span> * contact the HBase master and region servers. The hbck#connect() method must<a name="line.194"></a>
+<span class="sourceLineNo">195</span> * first be called successfully. Much of the region consistency information<a name="line.195"></a>
+<span class="sourceLineNo">196</span> * is transient and less risky to repair.<a name="line.196"></a>
+<span class="sourceLineNo">197</span> * <p><a name="line.197"></a>
+<span class="sourceLineNo">198</span> * If hbck is run from the command line, there are a handful of arguments that<a name="line.198"></a>
+<span class="sourceLineNo">199</span> * can be used to limit the kinds of repairs hbck will do. See the code in<a name="line.199"></a>
+<span class="sourceLineNo">200</span> * {@link #printUsageAndExit()} for more details.<a name="line.200"></a>
+<span class="sourceLineNo">201</span> */<a name="line.201"></a>
+<span class="sourceLineNo">202</span>@InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.TOOLS)<a name="line.202"></a>
+<span class="sourceLineNo">203</span>@InterfaceStability.Evolving<a name="line.203"></a>
+<span class="sourceLineNo">204</span>public class HBaseFsck extends Configured implements Closeable {<a name="line.204"></a>
+<span class="sourceLineNo">205</span> public static final long DEFAULT_TIME_LAG = 60000; // default value of 1 minute<a name="line.205"></a>
+<span class="sourceLineNo">206</span> public static final long DEFAULT_SLEEP_BEFORE_RERUN = 10000;<a name="line.206"></a>
+<span class="sourceLineNo">207</span> private static final int MAX_NUM_THREADS = 50; // #threads to contact regions<a name="line.207"></a>
+<span class="sourceLineNo">208</span> private static boolean rsSupportsOffline = true;<a name="line.208"></a>
+<span class="sourceLineNo">209</span> private static final int DEFAULT_OVERLAPS_TO_SIDELINE = 2;<a name="line.209"></a>
+<span class="sourceLineNo">210</span> private static final int DEFAULT_MAX_MERGE = 5;<a name="line.210"></a>
+<span class="sourceLineNo">211</span> private static final String TO_BE_LOADED = "to_be_loaded";<a name="line.211"></a>
+<span class="sourceLineNo">212</span> private static final String HBCK_LOCK_FILE = "hbase-hbck.lock";<a name="line.212"></a>
+<span class="sourceLineNo">213</span> private static final int DEFAULT_MAX_LOCK_FILE_ATTEMPTS = 5;<a name="line.213"></a>
+<span class="sourceLineNo">214</span> private static final int DEFAULT_LOCK_FILE_ATTEMPT_SLEEP_INTERVAL = 200; // milliseconds<a name="line.214"></a>
+<span class="sourceLineNo">215</span> private static final int DEFAULT_LOCK_FILE_ATTEMPT_MAX_SLEEP_TIME = 5000; // milliseconds<a name="line.215"></a>
+<span class="sourceLineNo">216</span> // We have to set the timeout value > HdfsConstants.LEASE_SOFTLIMIT_PERIOD.<a name="line.216"></a>
+<span class="sourceLineNo">217</span> // In HADOOP-2.6 and later, the Namenode proxy now created with custom RetryPolicy for<a name="line.217"></a>
+<span class="sourceLineNo">218</span> // AlreadyBeingCreatedException which is implies timeout on this operations up to<a name="line.218"></a>
+<span class="sourceLineNo">219</span> // HdfsConstants.LEASE_SOFTLIMIT_PERIOD (60 seconds).<a name="line.219"></a>
+<span class="sourceLineNo">220</span> private static final int DEFAULT_WAIT_FOR_LOCK_TIMEOUT = 80; // seconds<a name="line.220"></a>
+<span class="sourceLineNo">221</span> private static final int DEFAULT_MAX_CREATE_ZNODE_ATTEMPTS = 5;<a name="line.221"></a>
+<span class="sourceLineNo">222</span> private static final int DEFAULT_CREATE_ZNODE_ATTEMPT_SLEEP_INTERVAL = 200; // milliseconds<a name="line.222"></a>
+<span class="sourceLineNo">223</span> private static final int DEFAULT_CREATE_ZNODE_ATTEMPT_MAX_SLEEP_TIME = 5000; // milliseconds<a name="line.223"></a>
+<span class="sourceLineNo">224</span><a name="line.224"></a>
+<span class="sourceLineNo">225</span> /**********************<a name="line.225"></a>
+<span class="sourceLineNo">226</span> * Internal resources<a name="line.226"></a>
+<span class="sourceLineNo">227</span> **********************/<a name="line.227"></a>
+<span class="sourceLineNo">228</span> private static final Logger LOG = LoggerFactory.getLogger(HBaseFsck.class.getName());<a name="line.228"></a>
+<span class="sourceLineNo">229</span> private ClusterMetrics status;<a name="line.229"></a>
+<span class="sourceLineNo">230</span> private ClusterConnection connection;<a name="line.230"></a>
+<span class="sourceLineNo">231</span> private Admin admin;<a name="line.231"></a>
+<span class="sourceLineNo">232</span> private Table meta;<a name="line.232"></a>
+<span class="sourceLineNo">233</span> // threads to do ||izable tasks: retrieve data from regionservers, handle overlapping regions<a name="line.233"></a>
+<span class="sourceLineNo">234</span> protected ExecutorService executor;<a name="line.234"></a>
+<span class="sourceLineNo">235</span> private long startMillis = EnvironmentEdgeManager.currentTime();<a name="line.235"></a>
+<span class="sourceLineNo">236</span> private HFileCorruptionChecker hfcc;<a name="line.236"></a>
+<span class="sourceLineNo">237</span> private int retcode = 0;<a name="line.237"></a>
+<span class="sourceLineNo">238</span> private Path HBCK_LOCK_PATH;<a name="line.238"></a>
+<span class="sourceLineNo">239</span> private FSDataOutputStream hbckOutFd;<a name="line.239"></a>
+<span class="sourceLineNo">240</span> // This lock is to prevent cleanup of balancer resources twice between<a name="line.240"></a>
+<span class="sourceLineNo">241</span> // ShutdownHook and the main code. We cleanup only if the connect() is<a name="line.241"></a>
+<span class="sourceLineNo">242</span> // successful<a name="line.242"></a>
+<span class="sourceLineNo">243</span> private final AtomicBoolean hbckLockCleanup = new AtomicBoolean(false);<a name="line.243"></a>
+<span class="sourceLineNo">244</span><a name="line.244"></a>
+<span class="sourceLineNo">245</span> // Unsupported options in HBase 2.0+<a name="line.245"></a>
+<span class="sourceLineNo">246</span> private static final Set<String> unsupportedOptionsInV2 = Sets.newHashSet("-fix",<a name="line.246"></a>
+<span class="sourceLineNo">247</span> "-fixAssignments", "-fixMeta", "-fixHdfsHoles", "-fixHdfsOrphans", "-fixTableOrphans",<a name="line.247"></a>
+<span class="sourceLineNo">248</span> "-fixHdfsOverlaps", "-sidelineBigOverlaps", "-fixSplitParents", "-removeParents",<a name="line.248"></a>
+<span class="sourceLineNo">249</span> "-fixEmptyMetaCells", "-repair", "-repairHoles", "-maxOverlapsToSideline", "-maxMerge");<a name="line.249"></a>
+<span class="sourceLineNo">250</span><a name="line.250"></a>
+<span class="sourceLineNo">251</span> /***********<a name="line.251"></a>
+<span class="sourceLineNo">252</span> * Options<a name="line.252"></a>
+<span class="sourceLineNo">253</span> ***********/<a name="line.253"></a>
+<span class="sourceLineNo">254</span> private static boolean details = false; // do we display the full report<a name="line.254"></a>
+<span class="sourceLineNo">255</span> private long timelag = DEFAULT_TIME_LAG; // tables whose modtime is older<a name="line.255"></a>
+<span class="sourceLineNo">256</span> private static boolean forceExclusive = false; // only this hbck can modify HBase<a name="line.256"></a>
+<span class="sourceLineNo">257</span> private boolean fixAssignments = false; // fix assignment errors?<a name="line.257"></a>
+<span class="sourceLineNo">258</span> private boolean fixMeta = false; // fix meta errors?<a name="line.258"></a>
+<span class="sourceLineNo">259</span> private boolean checkHdfs = true; // load and check fs consistency?<a name="line.259"></a>
+<span class="sourceLineNo">260</span> private boolean fixHdfsHoles = false; // fix fs holes?<a name="line.260"></a>
+<span class="sourceLineNo">261</span> private boolean fixHdfsOverlaps = false; // fix fs overlaps (risky)<a name="line.261"></a>
+<span class="sourceLineNo">262</span> private boolean fixHdfsOrphans = false; // fix fs holes (missing .regioninfo)<a name="line.262"></a>
+<span class="sourceLineNo">263</span> private boolean fixTableOrphans = false; // fix fs holes (missing .tableinfo)<a name="line.263"></a>
+<span class="sourceLineNo">264</span> private boolean fixVersionFile = false; // fix missing hbase.version file in hdfs<a name="line.264"></a>
+<span class="sourceLineNo">265</span> private boolean fixSplitParents = false; // fix lingering split parents<a name="line.265"></a>
+<span class="sourceLineNo">266</span> private boolean removeParents = false; // remove split parents<a name="line.266"></a>
+<span class="sourceLineNo">267</span> private boolean fixReferenceFiles = false; // fix lingering reference store file<a name="line.267"></a>
+<span class="sourceLineNo">268</span> private boolean fixHFileLinks = false; // fix lingering HFileLinks<a name="line.268"></a>
+<span class="sourceLineNo">269</span> private boolean fixEmptyMetaCells = false; // fix (remove) empty REGIONINFO_QUALIFIER rows<a name="line.269"></a>
+<span class="sourceLineNo">270</span> private boolean fixReplication = false; // fix undeleted replication queues for removed peer<a name="line.270"></a>
+<span class="sourceLineNo">271</span> private boolean fixAny = false; // Set to true if any of the fix is required.<a name="line.271"></a>
+<span class="sourceLineNo">272</span><a name="line.272"></a>
+<span class="sourceLineNo">273</span> // limit checking/fixes to listed tables, if empty attempt to check/fix all<a name="line.273"></a>
+<span class="sourceLineNo">274</span> // hbase:meta are always checked<a name="line.274"></a>
+<span class="sourceLineNo">275</span> private Set<TableName> tablesIncluded = new HashSet<>();<a name="line.275"></a>
+<span class="sourceLineNo">276</span> private int maxMerge = DEFAULT_MAX_MERGE; // maximum number of overlapping regions to merge<a name="line.276"></a>
+<span class="sourceLineNo">277</span> // maximum number of overlapping regions to sideline<a name="line.277"></a>
+<span class="sourceLineNo">278</span> private int maxOverlapsToSideline = DEFAULT_OVERLAPS_TO_SIDELINE;<a name="line.278"></a>
+<span class="sourceLineNo">279</span> private boolean sidelineBigOverlaps = false; // sideline overlaps with >maxMerge regions<a name="line.279"></a>
+<span class="sourceLineNo">280</span> private Path sidelineDir = null;<a name="line.280"></a>
+<span class="sourceLineNo">281</span><a name="line.281"></a>
+<span class="sourceLineNo">282</span> private boolean rerun = false; // if we tried to fix something, rerun hbck<a name="line.282"></a>
+<span class="sourceLineNo">283</span> private static boolean summary = false; // if we want to print less output<a name="line.283"></a>
+<span class="sourceLineNo">284</span> private boolean checkMetaOnly = false;<a name="line.284"></a>
+<span class="sourceLineNo">285</span> private boolean checkRegionBoundaries = false;<a name="line.285"></a>
+<span class="sourceLineNo">286</span> private boolean ignorePreCheckPermission = false; // if pre-check permission<a name="line.286"></a>
+<span class="sourceLineNo">287</span><a name="line.287"></a>
+<span class="sourceLineNo">288</span> /*********<a name="line.288"></a>
+<span class="sourceLineNo">289</span> * State<a name="line.289"></a>
+<span class="sourceLineNo">290</span> *********/<a name="line.290"></a>
+<span class="sourceLineNo">291</span> final private ErrorReporter errors;<a name="line.291"></a>
+<span class="sourceLineNo">292</span> int fixes = 0;<a name="line.292"></a>
+<span class="sourceLineNo">293</span><a name="line.293"></a>
+<span class="sourceLineNo">294</span> /**<a name="line.294"></a>
+<span class="sourceLineNo">295</span> * This map contains the state of all hbck items. It maps from encoded region<a name="line.295"></a>
+<span class="sourceLineNo">296</span> * name to HbckInfo structure. The information contained in HbckInfo is used<a name="line.296"></a>
+<span class="sourceLineNo">297</span> * to detect and correct consistency (hdfs/meta/deployment) problems.<a name="line.297"></a>
+<span class="sourceLineNo">298</span> */<a name="line.298"></a>
+<span class="sourceLineNo">299</span> private TreeMap<String, HbckInfo> regionInfoMap = new TreeMap<>();<a name="line.299"></a>
+<span class="sourceLineNo">300</span> // Empty regioninfo qualifiers in hbase:meta<a name="line.300"></a>
+<span class="sourceLineNo">301</span> private Set<Result> emptyRegionInfoQualifiers = new HashSet<>();<a name="line.301"></a>
+<span class="sourceLineNo">302</span><a name="line.302"></a>
+<span class="sourceLineNo">303</span> /**<a name="line.303"></a>
+<span class="sourceLineNo">304</span> * This map from Tablename -> TableInfo contains the structures necessary to<a name="line.304"></a>
+<span class="sourceLineNo">305</span> * detect table consistency problems (holes, dupes, overlaps). It is sorted<a name="line.305"></a>
+<span class="sourceLineNo">306</span> * to prevent dupes.<a name="line.306"></a>
+<span class="sourceLineNo">307</span> *<a name="line.307"></a>
+<span class="sourceLineNo">308</span> * If tablesIncluded is empty, this map contains all tables.<a name="line.308"></a>
+<span class="sourceLineNo">309</span> * Otherwise, it contains only meta tables and tables in tablesIncluded,<a name="line.309"></a>
+<span class="sourceLineNo">310</span> * unless checkMetaOnly is specified, in which case, it contains only<a name="line.310"></a>
+<span class="sourceLineNo">311</span> * the meta table<a name="line.311"></a>
+<span class="sourceLineNo">312</span> */<a name="line.312"></a>
+<span class="sourceLineNo">313</span> private SortedMap<TableName, TableInfo> tablesInfo = new ConcurrentSkipListMap<>();<a name="line.313"></a>
+<span class="sourceLineNo">314</span><a name="line.314"></a>
+<span class="sourceLineNo">315</span> /**<a name="line.315"></a>
+<span class="sourceLineNo">316</span> * When initially looking at HDFS, we attempt to find any orphaned data.<a name="line.316"></a>
+<span class="sourceLineNo">317</span> */<a name="line.317"></a>
+<span class="sourceLineNo">318</span> private List<HbckInfo> orphanHdfsDirs = Collections.synchronizedList(new ArrayList<HbckInfo>());<a name="line.318"></a>
<span class="sourceLineNo">319</span><a name="line.319"></a>
-<span class="sourceLineNo">320</span> private ZKWatcher zkw = null;<a name="line.320"></a>
-<span class="sourceLineNo">321</span> private String hbckEphemeralNodePath = null;<a name="line.321"></a>
-<span class="sourceLineNo">322</span> private boolean hbckZodeCreated = false;<a name="line.322"></a>
-<span class="sourceLineNo">323</span><a name="line.323"></a>
-<span class="sourceLineNo">324</span> /**<a name="line.324"></a>
-<span class="sourceLineNo">325</span> * Constructor<a name="line.325"></a>
-<span class="sourceLineNo">326</span> *<a name="line.326"></a>
-<span class="sourceLineNo">327</span> * @param conf Configuration object<a name="line.327"></a>
-<span class="sourceLineNo">328</span> * @throws MasterNotRunningException if the master is not running<a name="line.328"></a>
-<span class="sourceLineNo">329</span> * @throws ZooKeeperConnectionException if unable to connect to ZooKeeper<a name="line.329"></a>
-<span class="sourceLineNo">330</span> */<a name="line.330"></a>
-<span class="sourceLineNo">331</span> public HBaseFsck(Configuration conf) throws IOException, ClassNotFoundException {<a name="line.331"></a>
-<span class="sourceLineNo">332</span> this(conf, createThreadPool(conf));<a name="line.332"></a>
-<span class="sourceLineNo">333</span> }<a name="line.333"></a>
-<span class="sourceLineNo">334</span><a name="line.334"></a>
-<span class="sourceLineNo">335</span> private static ExecutorService createThreadPool(Configuration conf) {<a name="line.335"></a>
-<span class="sourceLineNo">336</span> int numThreads = conf.getInt("hbasefsck.numthreads", MAX_NUM_THREADS);<a name="line.336"></a>
-<span class="sourceLineNo">337</span> return new ScheduledThreadPoolExecutor(numThreads, Threads.newDaemonThreadFactory("hbasefsck"));<a name="line.337"></a>
-<span class="sourceLineNo">338</span> }<a name="line.338"></a>
-<span class="sourceLineNo">339</span><a name="line.339"></a>
-<span class="sourceLineNo">340</span> /**<a name="line.340"></a>
-<span class="sourceLineNo">341</span> * Constructor<a name="line.341"></a>
-<span class="sourceLineNo">342</span> *<a name="line.342"></a>
-<span class="sourceLineNo">343</span> * @param conf<a name="line.343"></a>
-<span class="sourceLineNo">344</span> * Configuration object<a name="line.344"></a>
-<span class="sourceLineNo">345</span> * @throws MasterNotRunningException<a name="line.345"></a>
-<span class="sourceLineNo">346</span> * if the master is not running<a name="line.346"></a>
-<span class="sourceLineNo">347</span> * @throws ZooKeeperConnectionException<a name="line.347"></a>
-<span class="sourceLineNo">348</span> * if unable to connect to ZooKeeper<a name="line.348"></a>
-<span class="sourceLineNo">349</span> */<a name="line.349"></a>
-<span class="sourceLineNo">350</span> public HBaseFsck(Configuration conf, ExecutorService exec) throws MasterNotRunningException,<a name="line.350"></a>
-<span class="sourceLineNo">351</span> ZooKeeperConnectionException, IOException, ClassNotFoundException {<a name="line.351"></a>
-<span class="sourceLineNo">352</span> super(conf);<a name="line.352"></a>
-<span class="sourceLineNo">353</span> errors = getErrorReporter(getConf());<a name="line.353"></a>
-<span class="sourceLineNo">354</span> this.executor = exec;<a name="line.354"></a>
-<span class="sourceLineNo">355</span> lockFileRetryCounterFactory = new RetryCounterFactory(<a name="line.355"></a>
-<span class="sourceLineNo">356</span> getConf().getInt("hbase.hbck.lockfile.attempts", DEFAULT_MAX_LOCK_FILE_ATTEMPTS),<a name="line.356"></a>
-<span class="sourceLineNo">357</span> getConf().getInt(<a name="line.357"></a>
-<span class="sourceLineNo">358</span> "hbase.hbck.lockfile.attempt.sleep.interval", DEFAULT_LOCK_FILE_ATTEMPT_SLEEP_INTERVAL),<a name="line.358"></a>
-<span class="sourceLineNo">359</span> getConf().getInt(<a name="line.359"></a>
-<span class="sourceLineNo">360</span> "hbase.hbck.lockfile.attempt.maxsleeptime", DEFAULT_LOCK_FILE_ATTEMPT_MAX_SLEEP_TIME));<a name="line.360"></a>
-<span class="sourceLineNo">361</span> createZNodeRetryCounterFactory = new RetryCounterFactory(<a name="line.361"></a>
-<span class="sourceLineNo">362</span> getConf().getInt("hbase.hbck.createznode.attempts", DEFAULT_MAX_CREATE_ZNODE_ATTEMPTS),<a name="line.362"></a>
-<span class="sourceLineNo">363</span> getConf().getInt(<a name="line.363"></a>
-<span class="sourceLineNo">364</span> "hbase.hbck.createznode.attempt.sleep.interval",<a name="line.364"></a>
-<span class="sourceLineNo">365</span> DEFAULT_CREATE_ZNODE_ATTEMPT_SLEEP_INTERVAL),<a name="line.365"></a>
+<span class="sourceLineNo">320</span> private Map<TableName, Set<String>> orphanTableDirs = new HashMap<>();<a name="line.320"></a>
+<span class="sourceLineNo">321</span> private Map<TableName, TableState> tableStates = new HashMap<>();<a name="line.321"></a>
+<span class="sourceLineNo">322</span> private final RetryCounterFactory lockFileRetryCounterFactory;<a name="line.322"></a>
+<span class="sourceLineNo">323</span> private final RetryCounterFactory createZNodeRetryCounterFactory;<a name="line.323"></a>
+<span class="sourceLineNo">324</span><a name="line.324"></a>
+<span class="sourceLineNo">325</span> private Map<TableName, Set<String>> skippedRegions = new HashMap<>();<a name="line.325"></a>
+<span class="sourceLineNo">326</span><a name="line.326"></a>
+<span class="sourceLineNo">327</span> private ZKWatcher zkw = null;<a name="line.327"></a>
+<span class="sourceLineNo">328</span> private String hbckEphemeralNodePath = null;<a name="line.328"></a>
+<span class="sourceLineNo">329</span> private boolean hbckZodeCreated = false;<a name="line.329"></a>
+<span class="sourceLineNo">330</span><a name="line.330"></a>
+<span class="sourceLineNo">331</span> /**<a name="line.331"></a>
+<span class="sourceLineNo">332</span> * Constructor<a name="line.332"></a>
+<span class="sourceLineNo">333</span> *<a name="line.333"></a>
+<span class="sourceLineNo">334</span> * @param conf Configuration object<a name="line.334"></a>
+<span class="sourceLineNo">335</span> * @throws MasterNotRunningException if the master is not running<a name="line.335"></a>
+<span class="sourceLineNo">336</span> * @throws ZooKeeperConnectionException if unable to connect to ZooKeeper<a name="line.336"></a>
+<span class="sourceLineNo">337</span> */<a name="line.337"></a>
+<span class="sourceLineNo">338</span> public HBaseFsck(Configuration conf) throws IOException, ClassNotFoundException {<a name="line.338"></a>
+<span class="sourceLineNo">339</span> this(conf, createThreadPool(conf));<a name="line.339"></a>
+<span class="sourceLineNo">340</span> }<a name="line.340"></a>
+<span class="sourceLineNo">341</span><a name="line.341"></a>
+<span class="sourceLineNo">342</span> private static ExecutorService createThreadPool(Configuration conf) {<a name="line.342"></a>
+<span class="sourceLineNo">343</span> int numThreads = conf.getInt("hbasefsck.numthreads", MAX_NUM_THREADS);<a name="line.343"></a>
+<span class="sourceLineNo">344</span> return new ScheduledThreadPoolExecutor(numThreads, Threads.newDaemonThreadFactory("hbasefsck"));<a name="line.344"></a>
+<span class="sourceLineNo">345</span> }<a name="line.345"></a>
+<span class="sourceLineNo">346</span><a name="line.346"></a>
+<span class="sourceLineNo">347</span> /**<a name="line.347"></a>
+<span class="sourceLineNo">348</span> * Constructor<a name="line.348"></a>
+<span class="sourceLineNo">349</span> *<a name="line.349"></a>
+<span class="sourceLineNo">350</span> * @param conf<a name="line.350"></a>
+<span class="sourceLineNo">351</span> * Configuration object<a name="line.351"></a>
+<span class="sourceLineNo">352</span> * @throws MasterNotRunningException<a name="line.352"></a>
+<span class="sourceLineNo">353</span> * if the master is not running<a name="line.353"></a>
+<span class="sourceLineNo">354</span> * @throws ZooKeeperConnectionException<a name="line.354"></a>
+<span class="sourceLineNo">355</span> * if unable to connect to ZooKeeper<a name="line.355"></a>
+<span class="sourceLineNo">356</span> */<a name="line.356"></a>
+<span class="sourceLineNo">357</span> public HBaseFsck(Configuration conf, ExecutorService exec) throws MasterNotRunningException,<a name="line.357"></a>
+<span class="sourceLineNo">358</span> ZooKeeperConnectionException, IOException, ClassNotFoundException {<a name="line.358"></a>
+<span class="sourceLineNo">359</span> super(conf);<a name="line.359"></a>
+<span class="sourceLineNo">360</span> errors = getErrorReporter(getConf());<a name="line.360"></a>
+<span class="sourceLineNo">361</span> this.executor = exec;<a name="line.361"></a>
+<span class="sourceLineNo">362</span> lockFileRetryCounterFactory = new RetryCounterFactory(<a name="line.362"></a>
+<span class="sourceLineNo">363</span> getConf().getInt("hbase.hbck.lockfile.attempts", DEFAULT_MAX_LOCK_FILE_ATTEMPTS),<a name="line.363"></a>
+<span class="sourceLineNo">364</span> getConf().getInt(<a name="line.364"></a>
+<span class="sourceLineNo">365</span> "hbase.hbck.lockfile.attempt.sleep.interval", DEFAULT_LOCK_FILE_ATTEMPT_SLEEP_INTERVAL),<a name="line.365"></a>
<span class="sourceLineNo">366</span> getConf().getInt(<a name="line.366"></a>
-<span class="sourceLineNo">367</span> "hbase.hbck.createznode.attempt.maxsleeptime",<a name="line.367"></a>
-<span class="sourceLineNo">368</span> DEFAULT_CREATE_ZNODE_ATTEMPT_MAX_SLEEP_TIME));<a name="line.368"></a>
-<span class="sourceLineNo">369</span> zkw = createZooKeeperWatcher();<a name="line.369"></a>
-<span class="sourceLineNo">370</span> }<a name="line.370"></a>
-<span class="sourceLineNo">371</span><a name="line.371"></a>
-<span class="sourceLineNo">372</span> private class FileLockCallable implements Callable<FSDataOutputStream> {<a name="line.372"></a>
-<span class="sourceLineNo">373</span> RetryCounter retryCounter;<a name="line.373"></a>
-<span class="sourceLineNo">374</span><a name="line.374"></a>
-<span class="sourceLineNo">375</span> public FileLockCallable(RetryCounter retryCounter) {<a name="line.375"></a>
-<span class="sourceLineNo">376</span> this.retryCounter = retryCounter;<a name="line.376"></a>
-<span class="sourceLineNo">377</span> }<a name="line.377"></a>
-<span class="sourceLineNo">378</span> @Override<a name="line.378"></a>
-<span class="sourceLineNo">379</span> public FSDataOutputStream call() throws IOException {<a name="line.379"></a>
-<span class="sourceLineNo">380</span> try {<a name="line.380"></a>
-<span class="sourceLineNo">381</span> FileSystem fs = FSUtils.getCurrentFileSystem(getConf());<a name="line.381"></a>
-<span class="sourceLineNo">382</span> FsPermission defaultPerms = FSUtils.getFilePermissions(fs, getConf(),<a name="line.382"></a>
-<span class="sourceLineNo">383</span> HConstants.DATA_FILE_UMASK_KEY);<a name="line.383"></a>
-<span class="sourceLineNo">384</span> Path tmpDir = new Path(FSUtils.getRootDir(getConf()), HConstants.HBASE_TEMP_DIRECTORY);<a name="line.384"></a>
-<span class="sourceLineNo">385</span> fs.mkdirs(tmpDir);<a name="line.385"></a>
-<span class="sourceLineNo">386</span> HBCK_LOCK_PATH = new Path(tmpDir, HBCK_LOCK_FILE);<a name="line.386"></a>
-<span class="sourceLineNo">387</span> final FSDataOutputStream out = createFileWithRetries(fs, HBCK_LOCK_PATH, defaultPerms);<a name="line.387"></a>
-<span class="sourceLineNo">388</span> out.writeBytes(InetAddress.getLocalHost().toString());<a name="line.388"></a>
-<span class="sourceLineNo">389</span> out.flush();<a name="line.389"></a>
-<span class="sourceLineNo">390</span> return out;<a name="line.390"></a>
-<span class="sourceLineNo">391</span> } catch(RemoteException e) {<a name="line.391"></a>
-<span class="sourceLineNo">392</span> if(AlreadyBeingCreatedException.class.getName().equals(e.getClassName())){<a name="line.392"></a>
-<span class="sourceLineNo">393</span> return null;<a name="line.393"></a>
-<span class="sourceLineNo">394</span> } else {<a name="line.394"></a>
-<span class="sourceLineNo">395</span> throw e;<a name="line.395"></a>
-<span class="sourceLineNo">396</span> }<a name="line.396"></a>
-<span class="sourceLineNo">397</span> }<a name="line.397"></a>
-<span class="sourceLineNo">398</span> }<a name="line.398"></a>
-<span class="sourceLineNo">399</span><a name="line.399"></a>
-<span class="sourceLineNo">400</span> private FSDataOutputStream createFileWithRetries(final FileSystem fs,<a name="line.400"></a>
-<span class="sourceLineNo">401</span> final Path hbckLockFilePath, final FsPermission defaultPerms)<a name="line.401"></a>
-<span class="sourceLineNo">402</span> throws IOException {<a name="line.402"></a>
-<span class="sourceLineNo">403</span><a name="line.403"></a>
-<span class="sourceLineNo">404</span> IOException exception = null;<a name="line.404"></a>
-<span class="sourceLineNo">405</span> do {<a name="line.405"></a>
-<span class="sourceLineNo">406</span> try {<a name="line.406"></a>
-<span class="sourceLineNo">407</span> return FSUtils.create(fs, hbckLockFilePath, defaultPerms, false);<a name="line.407"></a>
-<span class="sourceLineNo">408</span> } catch (IOException ioe) {<a name="line.408"></a>
-<span class="sourceLineNo">409</span> LOG.info("Failed to create lock file " + hbckLockFilePath.getName()<a name="line.409"></a>
-<span class="sourceLineNo">410</span> + ", try=" + (retryCounter.getAttemptTimes() + 1) + " of "<a name="line.410"></a>
-<span class="sourceLineNo">411</span> + retryCounter.getMaxAttempts());<a name="line.411"></a>
-<span class="sourceLineNo">412</span> LOG.debug("Failed to create lock file " + hbckLockFilePath.getName(),<a name="line.412"></a>
-<span class="sourceLineNo">413</span> ioe);<a name="line.413"></a>
-<span class="sourceLineNo">414</span> try {<a name="line.414"></a>
-<span class="sourceLineNo">415</span> exception = ioe;<a name="line.415"></a>
-<span class="sourceLineNo">416</span> retryCounter.sleepUntilNextRetry();<a name="line.416"></a>
-<span class="sourceLineNo">417</span> } catch (InterruptedException ie) {<a name="line.417"></a>
-<span class="sourceLineNo">418</span> throw (InterruptedIOException) new InterruptedIOException(<a name="line.418"></a>
-<span class="sourceLineNo">419</span> "Can't create lock file " + hbckLockFilePath.getName())<a name="line.419"></a>
-<span class="sourceLineNo">420</span> .initCause(ie);<a name="line.420"></a>
-<span class="sourceLineNo">421</span> }<a name="line.421"></a>
-<span class="sourceLineNo">422</span> }<a name="line.422"></a>
-<span class="sourceLineNo">423</span> } while (retryCounter.shouldRetry());<a name="line.423"></a>
-<span class="sourceLineNo">424</span><a name="line.424"></a>
-<span class="sourceLineNo">425</span> throw exception;<a name="line.425"></a>
-<span class="sourceLineNo">426</span> }<a name="line.426"></a>
-<span class="sourceLineNo">427</span> }<a name="line.427"></a>
-<span class="sourceLineNo">428</span><a name="line.428"></a>
-<span class="sourceLineNo">429</span> /**<a name="line.429"></a>
-<span class="sourceLineNo">430</span> * This method maintains a lock using a file. If the creation fails we return null<a name="line.430"></a>
-<span class="sourceLineNo">431</span> *<a name="line.431"></a>
-<span class="sourceLineNo">432</span> * @return FSDataOutputStream object corresponding to the newly opened lock file<a name="line.432"></a>
-<span class="sourceLineNo">433</span> * @throws IOException if IO failure occurs<a name="line.433"></a>
-<span class="sourceLineNo">434</span> */<a name="line.434"></a>
-<span class="sourceLineNo">435</span> private FSDataOutputStream checkAndMarkRunningHbck() throws IOException {<a name="line.435"></a>
-<span class="sourceLineNo">436</span> RetryCounter retryCounter = lockFileRetryCounterFactory.create();<a name="line.436"></a>
-<span class="sourceLineNo">437</span> FileLockCallable callable = new FileLockCallable(retryCounter);<a name="line.437"></a>
-<span class="sourceLineNo">438</span> ExecutorService executor = Executors.newFixedThreadPool(1);<a name="line.438"></a>
-<span class="sourceLineNo">439</span> FutureTask<FSDataOutputStream> futureTask = new FutureTask<>(callable);<a name="line.439"></a>
-<span class="sourceLineNo">440</span> executor.execute(futureTask);<a name="line.440"></a>
-<span class="sourceLineNo">441</span> final int timeoutInSeconds = getConf().getInt(<a name="line.441"></a>
-<span class="sourceLineNo">442</span> "hbase.hbck.lockfile.maxwaittime", DEFAULT_WAIT_FOR_LOCK_TIMEOUT);<a name="line.442"></a>
-<span class="sourceLineNo">443</span> FSDataOutputStream stream = null;<a name="line.443"></a>
-<span class="sourceLineNo">444</span> try {<a name="line.444"></a>
-<span class="sourceLineNo">445</span> stream = futureTask.get(timeoutInSeconds, TimeUnit.SECONDS);<a name="line.445"></a>
-<span class="sourceLineNo">446</span> } catch (ExecutionException ee) {<a name="line.446"></a>
-<span class="sourceLineNo">447</span> LOG.warn("Encountered exception when opening lock file", ee);<a name="line.447"></a>
-<span class="sourceLineNo">448</span> } catch (InterruptedException ie) {<a name="line.448"></a>
-<span class="sourceLineNo">449</span> LOG.warn("Interrupted when opening lock file", ie);<a name="line.449"></a>
-<span class="sourceLineNo">450</span> Thread.currentThread().interrupt();<a name="line.450"></a>
-<span class="sourceLineNo">451</span> } catch (TimeoutException exception) {<a name="line.451"></a>
-<span class="sourceLineNo">452</span> // took too long to obtain lock<a name="line.452"></a>
-<span class="sourceLineNo">453</span> LOG.warn("Took more than " + timeoutInSeconds + " seconds in obtaining lock");<a name="line.453"></a>
-<span class="sourceLineNo">454</span> futureTask.cancel(true);<a name="line.454"></a>
-<span class="sourceLineNo">455</span> } finally {<a name="line.455"></a>
-<span class="sourceLineNo">456</span> executor.shutdownNow();<a name="line.456"></a>
-<span class="sourceLineNo">457</span> }<a name="line.457"></a>
-<span class="sourceLineNo">458</span> return stream;<a name="line.458"></a>
-<span class="sourceLineNo">459</span> }<a name="line.459"></a>
-<span class="sourceLineNo">460</span><a name="line.460"></a>
-<span class="sourceLineNo">461</span> private void unlockHbck() {<a name="line.461"></a>
-<span class="sourceLineNo">462</span> if (isExclusive() && hbckLockCleanup.compareAndSet(true, false)) {<a name="line.462"></a>
-<span class="sourceLineNo">463</span> RetryCounter retryCounter = lockFileRetryCounterFactory.create();<a name="line.463"></a>
-<span class="sourceLineNo">464</span> do {<a name="line.464"></a>
-<span class="sourceLineNo">465</span> try {<a name="line.465"></a>
-<span class="sourceLineNo">466</span> IOUtils.closeQuietly(hbckOutFd);<a name="line.466"></a>
-<span class="sourceLineNo">467</span> FSUtils.delete(FSUtils.getCurrentFileSystem(getConf()),<a name="line.467"></a>
-<span class="sourceLineNo">468</span> HBCK_LOCK_PATH, true);<a name="line.468"></a>
-<span class="sourceLineNo">469</span> LOG.info("Finishing hbck");<a name="line.469"></a>
-<span class="sourceLineNo">470</span> return;<a name="line.470"></a>
-<span class="sourceLineNo">471</span> } catch (IOException ioe) {<a name="line.471"></a>
-<span class="sourceLineNo">472</span> LOG.info("Failed to delete " + HBCK_LOCK_PATH + ", try="<a name="line.472"></a>
-<span class="sourceLineNo">473</span> + (retryCounter.getAttemptTimes() + 1) + " of "<a name="line.473"></a>
-<span class="sourceLineNo">474</span> + retryCounter.getMaxAttempts());<a name="line.474"></a>
-<span class="sourceLineNo">475</span> LOG.debug("Failed to delete " + HBCK_LOCK_PATH, ioe);<a name="line.475"></a>
-<span class="sourceLineNo">476</span> try {<a name="line.476"></a>
-<span class="sourceLineNo">477</span> retryCounter.sleepUntilNextRetry();<a name="line.477"></a>
-<span class="sourceLineNo">478</span> } catch (InterruptedException ie) {<a name="line.478"></a>
-<span class="sourceLineNo">479</span> Thread.currentThread().interrupt();<a name="line.479"></a>
-<span class="sourceLineNo">480</span> LOG.warn("Interrupted while deleting lock file" +<a name="line.480"></a>
-<span class="sourceLineNo">481</span> HBCK_LOCK_PATH);<a name="line.481"></a>
-<span class="sourceLineNo">482</span> return;<a name="line.482"></a>
-<span class="sourceLineNo">483</span> }<a name="line.483"></a>
-<span class="sourceLineNo">484</span> }<a name="line.484"></a>
-<span class="sourceLineNo">485</span> } while (retryCounter.shouldRetry());<a name="line.485"></a>
-<span class="sourceLineNo">486</span> }<a name="line.486"></a>
-<span class="sourceLineNo">487</span> }<a name="line.487"></a>
-<span class="sourceLineNo">488</span><a name="line.488"></a>
-<span class="sourceLineNo">489</span> /**<a name="line.489"></a>
-<span class="sourceLineNo">490</span> * To repair region consistency, one must call connect() in order to repair<a name="line.490"></a>
-<span class="sourceLineNo">491</span> * online state.<a name="line.491"></a>
-<span class="sourceLineNo">492</span> */<a name="line.492"></a>
-<span class="sourceLineNo">493</span> public void connect() throws IOException {<a name="line.493"></a>
-<span class="sourceLineNo">494</span><a name="line.494"></a>
-<span class="sourceLineNo">495</span> if (isExclusive()) {<a name="line.495"></a>
-<span class="sourceLineNo">496</span> // Grab the lock<a name="line.496"></a>
-<span class="sourceLineNo">497</span> hbckOutFd = checkAndMarkRunningHbck();<a name="line.497"></a>
-<span class="sourceLineNo">498</span> if (hbckOutFd == null) {<a name="line.498"></a>
-<span class="sourceLineNo">499</span> setRetCode(-1);<a name="line.499"></a>
-<span class="sourceLineNo">500</span> LOG.error("Another instance of hbck is fixing HBase, exiting this instance. " +<a name="line.500"></a>
-<span class="sourceLineNo">501</span> "[If you are sure no other instance is running, delete the lock file " +<a name="line.501"></a>
-<span class="sourceLineNo">502</span> HBCK_LOCK_PATH + " and rerun the tool]");<a name="line.502"></a>
-<span class="sourceLineNo">503</span> throw new IOException("Duplicate hbck - Abort");<a name="line.503"></a>
-<span class="sourceLineNo">504</span> }<a name="line.504"></a>
-<span class="sourceLineNo">505</span><a name="line.505"></a>
-<span class="sourceLineNo">506</span> // Make sure to cleanup the lock<a name="line.506"></a>
-<span class="sourceLineNo">507</span> hbckLockCleanup.set(true);<a name="line.507"></a>
-<span class="sourceLineNo">508</span> }<a name="line.508"></a>
-<span class="sourceLineNo">509</span><a name="line.509"></a>
-<span class="sourceLineNo">510</span><a name="line.510"></a>
-<span class="sourceLineNo">511</span> // Add a shutdown hook to this thread, in case user tries to<a name="line.511"></a>
-<span class="sourceLineNo">512</span> // kill the hbck with a ctrl-c, we want to cleanup the lock so that<a name="line.512"></a>
-<span class="sourceLineNo">513</span> // it is available for further calls<a name="line.513"></a>
-<span class="sourceLineNo">514</span> Runtime.getRuntime().addShutdownHook(new Thread() {<a name="line.514"></a>
-<span class="sourceLineNo">515</span> @Override<a name="line.515"></a>
-<span class="sourceLineNo">516</span> public void run() {<a name="line.516"></a>
-<span class="sourceLineNo">517</span> IOUtils.closeQuietly(HBaseFsck.this);<a name="line.517"></a>
-<span class="sourceLineNo">518</span> cleanupHbckZnode();<a name="line.518"></a>
-<span class="sourceLineNo">519</span> unlockHbck();<a name="line.519"></a>
-<span class="sourceLineNo">520</span> }<a name="line.520"></a>
-<span class="sourceLineNo">521</span> });<a name="line.521"></a>
-<span class="sourceLineNo">522</span><a name="line.522"></a>
-<span class="sourceLineNo">523</span> LOG.info("Launching hbck");<a name="line.523"></a>
-<span class="sourceLineNo">524</span><a name="line.524"></a>
-<span class="sourceLineNo">525</span> connection = (ClusterConnection)ConnectionFactory.createConnection(getConf());<a name="line.525"></a>
-<span class="sourceLineNo">526</span> admin = connection.getAdmin();<a name="line.526"></a>
-<span class="sourceLineNo">527</span> meta = connection.getTable(TableName.META_TABLE_NAME);<a name="line.527"></a>
-<span class="sourceLineNo">528</span> status = admin.getClusterMetrics(EnumSet.of(Option.LIVE_SERVERS,<a name="line.528"></a>
-<span class="sourceLineNo">529</span> Option.DEAD_SERVERS, Option.MASTER, Option.BACKUP_MASTERS,<a name="line.529"></a>
-<span class="sourceLineNo">530</span> Option.REGIONS_IN_TRANSITION, Option.HBASE_VERSION));<a name="line.530"></a>
-<span class="sourceLineNo">531</span> }<a name="line.531"></a>
-<span class="sourceLineNo">532</span><a name="line.532"></a>
-<span class="sourceLineNo">533</span> /**<a name="line.533"></a>
-<span class="sourceLineNo">534</span> * Get deployed regions according to the region servers.<a name="line.534"></a>
-<span class="sourceLineNo">535</span> */<a name="line.535"></a>
-<span class="sourceLineNo">536</span> private void loadDeployedRegions() throws IOException, InterruptedException {<a name="line.536"></a>
-<span class="sourceLineNo">537</span> // From the master, get a list of all known live region servers<a name="line.537"></a>
-<span class="sourceLineNo">538</span> Collection<ServerName> regionServers = status.getLiveServerMetrics().keySet();<a name="line.538"></a>
-<span class="sourceLineNo">539</span> errors.print("Number of live region servers: " + regionServers.size());<a name="line.539"></a>
-<span class="sourceLineNo">540</span> if (details) {<a name="line.540"></a>
-<span class="sourceLineNo">541</span> for (ServerName rsinfo: regionServers) {<a name="line.541"></a>
-<span class="sourceLineNo">542</span> errors.print(" " + rsinfo.getServerName());<a name="line.542"></a>
-<span class="sourceLineNo">543</span> }<a name="line.543"></a>
-<span class="sourceLineNo">544</span> }<a name="line.544"></a>
-<span class="sourceLineNo">545</span><a name="line.545"></a>
-<span class="sourceLineNo">546</span> // From the master, get a list of all dead region servers<a name="line.546"></a>
-<span class="sourceLineNo">547</span> Collection<ServerName> deadRegionServers = status.getDeadServerNames();<a name="line.547"></a>
-<span class="sourceLineNo">548</span> errors.print("Number of dead region servers: " + deadRegionServers.size());<a name="line.548"></a>
-<span class="sourceLineNo">549</span> if (details) {<a name="line.549"></a>
-<span class="sourceLineNo">550</span> for (ServerName name: deadRegionServers) {<a name="line.550"></a>
-<span class="sourceLineNo">551</span> errors.print(" " + name);<a name="line.551"></a>
-<span class="sourceLineNo">552</span> }<a name="line.552"></a>
-<span class="sourceLineNo">553</span> }<a name="line.553"></a>
-<span class="sourceLineNo">554</span><a name="line.554"></a>
-<span class="sourceLineNo">555</span> // Print the current master name and state<a name="line.555"></a>
-<span class="sourceLineNo">556</span> errors.print("Master: " + status.getMasterName());<a name="line.556"></a>
-<span class="sourceLineNo">557</span><a name="line.557"></a>
-<span class="sourceLineNo">558</span> // Print the list of all backup masters<a name="line.558"></a>
-<span class="sourceLineNo">559</span> Collection<ServerName> backupMasters = status.getBackupMasterNames();<a name="line.559"></a>
-<span class="sourceLineNo">560</span> errors.print("Number of backup masters: " + backupMasters.size());<a name="line.560"></a>
-<span class="sourceLineNo">561</span> if (details) {<a name="line.561"></a>
-<span class="sourceLineNo">562</span> for (ServerName name: backupMasters) {<a name="line.562"></a>
-<span class="sourceLineNo">563</span> errors.print(" " + name);<a name="line.563"></a>
-<span class="sourceLineNo">564</span> }<a name="line.564"></a>
-<span class="sourceLineNo">565</span> }<a name="line.565"></a>
-<span class="sourceLineNo">566</span><a name="line.566"></a>
-<span class="sourceLineNo">567</span> errors.print("Average load: " + status.getAverageLoad());<a name="line.567"></a>
-<span class="sourceLineNo">568</span> errors.print("Number of requests: " + status.getRequestCount());<a name="line.568"></a>
-<span class="sourceLineNo">569</span> errors.print("Number of regions: " + status.getRegionCount());<a name="line.569"></a>
-<span class="sourceLineNo">570</span><a name="line.570"></a>
-<span class="sourceLineNo">571</span> List<RegionState> rits = status.getRegionStatesInTransition();<a name="line.571"></a>
-<span class="sourceLineNo">572</span> errors.print("Number of regions in transition: " + rits.size());<a name="line.572"></a>
-<span class="sourceLineNo">573</span> if (details) {<a name="line.573"></a>
-<span class="sourceLineNo">574</span> for (RegionState state: rits) {<a name="line.574"></a>
-<span class="sourceLineNo">575</span> errors.print(" " + state.toDescriptiveString());<a name="line.575"></a>
-<span class="sourceLineNo">576</span> }<a name="line.576"></a>
-<span class="sourceLineNo">577</span> }<a name="line.577"></a>
-<span class="sourceLineNo">578</span><a name="line.578"></a>
-<span class="sourceLineNo">579</span> // Determine what's deployed<a name="line.579"></a>
-<span class="sourceLineNo">580</span> processRegionServers(regionServers);<a name="line.580"></a>
-<span class="sourceLineNo">581</span> }<a name="line.581"></a>
-<span class="sourceLineNo">582</span><a name="line.582"></a>
-<span class="sourceLineNo">583</span> /**<a name="line.583"></a>
-<span class="sourceLineNo">584</span> * Clear the current state of hbck.<a name="line.584"></a>
-<span class="sourceLineNo">585</span> */<a name="line.585"></a>
-<span class="sourceLineNo">586</span> private void clearState() {<a name="line.586"></a>
-<span class="sourceLineNo">587</span> // Make sure regionInfo is empty before starting<a name="line.587"></a>
-<span class="sourceLineNo">588</span> fixes = 0;<a name="line.588"></a>
-<span class="sourceLineNo">589</span> regionInfoMap.clear();<a name="line.589"></a>
-<span class="sourceLineNo">590</span> emptyRegionInfoQualifiers.clear();<a name="line.590"></a>
-<span class="sourceLineNo">591</span> tableStates.clear();<a name="line.591"></a>
-<span class="sourceLineNo">592</span> errors.clear();<a name="line.592"></a>
-<span class="sourceLineNo">593</span> tablesInfo.clear();<a name="line.593"></a>
-<span class="sourceLineNo">594</span> orphanHdfsDirs.clear();<a name="line.594"></a>
-<span class="sourceLineNo">595</span> skippedRegions.clear();<a name="line.595"></a>
-<span class="sourceLineNo">596</span> }<a name="line.596"></a>
-<span class="sourceLineNo">597</span><a name="line.597"></a>
-<span class="sourceLineNo">598</span> /**<a name="line.598"></a>
-<span class="sourceLineNo">599</span> * This repair method analyzes hbase data in hdfs and repairs it to satisfy<a name="line.599"></a>
-<span class="sourceLineNo">600</span> * the table integrity rules. HBase doesn't need to be online for this<a name="line.600"></a>
-<span class="sourceLineNo">601</span> * operation to work.<a name="line.601"></a>
-<span class="sourceLineNo">602</span> */<a name="line.602"></a>
-<span class="sourceLineNo">603</span> public void offlineHdfsIntegrityRepair() throws IOException, InterruptedException {<a name="line.603"></a>
-<span class="sourceLineNo">604</span> // Initial pass to fix orphans.<a name="line.604"></a>
-<span class="sourceLineNo">605</span> if (shouldCheckHdfs() && (shouldFixHdfsOrphans() || shouldFixHdfsHoles()<a name="line.605"></a>
-<span class="sourceLineNo">606</span> || shouldFixHdfsOverlaps() || shouldFixTableOrphans())) {<a name="line.606"></a>
-<span class="sourceLineNo">607</span> LOG.info("Loading regioninfos HDFS");<a name="line.607"></a>
-<span class="sourceLineNo">608</span> // if nothing is happening this should always complete in two iterations.<a name="line.608"></a>
-<span class="sourceLineNo">609</span> int maxIterations = getConf().getInt("hbase.hbck.integrityrepair.iterations.max", 3);<a name="line.609"></a>
-<span class="sourceLineNo">610</span> int curIter = 0;<a name="line.610"></a>
-<span class="sourceLineNo">611</span> do {<a name="line.611"></a>
-<span class="sourceLineNo">612</span> clearState(); // clears hbck state and reset fixes to 0 and.<a name="line.612"></a>
-<span class="sourceLineNo">613</span> // repair what's on HDFS<a name="line.613"></a>
-<span class="sourceLineNo">614</span> restoreHdfsIntegrity();<a name="line.614"></a>
-<span class="sourceLineNo">615</span> curIter++;// limit the number of iterations.<a name="line.615"></a>
-<span class="sourceLineNo">616</span> } while (fixes > 0 && curIter <= maxIterations);<a name="line.616"></a>
-<span class="sourceLineNo">617</span><a name="line.617"></a>
-<span class="sourceLineNo">618</span> // Repairs should be done in the first iteration and verification in the second.<a name="line.618"></a>
-<span class="sourceLineNo">619</span> // If there are more than 2 passes, something funny has happened.<a name="line.619"></a>
-<span class="sourceLineNo">620</span> if (curIter > 2) {<a name="line.620"></a>
-<span class="sourceLineNo">621</span> if (curIter == maxIterations) {<a name="line.621"></a>
-<span class="sourceLineNo">622</span> LOG.warn("Exiting integrity repairs after max " + curIter + " iterations. "<a name="line.622"></a>
-<span class="sourceLineNo">623</span> + "Tables integrity may not be fully repaired!");<a name="line.623"></a>
-<span class="sourceLineNo">624</span> } else {<a name="line.624"></a>
-<span class="sourceLineNo">625</span> LOG.info("Successfully exiting integrity repairs after " + curIter + " iterations");<a name="line.625"></a>
-<span class="sourceLineNo">626</span> }<a name="line.626"></a>
-<span class="sourceLineNo">627</span> }<a name="line.627"></a>
-<span class="sourceLineNo">628</span> }<a name="line.628"></a>
-<span class="sourceLineNo">629</span> }<a name="line.629"></a>
-<span class="sourceLineNo">630</span><a name="line.630"></a>
-<span class="sourceLineNo">631</span> /**<a name="line.631"></a>
-<span class="sourceLineNo">632</span> * This repair method requires the cluster to be online since it contacts<a name="line.632"></a>
-<span class="sourceLineNo">633</span> * region servers and the masters. It makes each region's state in HDFS, in<a name="line.633"></a>
-<span class="sourceLineNo">634</span> * hbase:meta, and deployments consistent.<a name="line.634"></a>
-<span class="sourceLineNo">635</span> *<a name="line.635"></a>
-<span class="sourceLineNo">636</span> * @return If &gt; 0 , number of errors detected, if &lt; 0 there was an unrecoverable<a name="line.636"></a>
-<span class="sourceLineNo">637</span> * error. If 0, we have a clean hbase.<a name="line.637"></a>
-<span class="sourceLineNo">638</span> */<a name="line.638"></a>
-<span class="sourceLineNo">639</span> public int onlineConsistencyRepair() throws IOException, KeeperException,<a name="line.639"></a>
-<span class="sourceLineNo">640</span> InterruptedException {<a name="line.640"></a>
-<span class="sourceLineNo">641</span><a name="line.641"></a>
-<span class="sourceLineNo">642</span> // get regions according to what is online on each RegionServer<a name="line.642"></a>
-<span class="sourceLineNo">643</span> loadDeployedRegions();<a name="line.643"></a>
-<span class="sourceLineNo">644</span> // check whether hbase:meta is deployed and online<a name="line.644"></a>
-<span class="sourceLineNo">645</span> recordMetaRegion();<a name="line.645"></a>
-<span class="sourceLineNo">646</span> // Check if hbase:meta is found only once and in the right place<a name="line.646"></a>
-<span class="sourceLineNo">647</span> if (!checkMetaRegion()) {<a name="line.647"></a>
-<span class="sourceLineNo">648</span> String errorMsg = "hbase:meta table is not consistent. ";<a name="line.648"></a>
-<span class="sourceLineNo">649</span> if (shouldFixAssignments()) {<a name="line.649"></a>
-<span class="sourceLineNo">650</span> errorMsg += "HBCK will try fixing it. Rerun once hbase:meta is back to consistent state.";<a name="line.650"></a>
-<span class="sourceLineNo">651</span> } else {<a name="line.651"></a>
-<span class="sourceLineNo">652</span> errorMsg += "Run HBCK with proper fix options to fix hbase:meta inconsistency.";<a name="line.652"></a>
-<span class="sourceLineNo">653</span> }<a name="line.653"></a>
-<span class="sourceLineNo">654</span> errors.reportError(errorMsg + " Exiting...");<a name="line.654"></a>
-<span class="sourceLineNo">655</span> return -2;<a name="line.655"></a>
-<span class="sourceLineNo">656</span> }<a name="line.656"></a>
-<span class="sourceLineNo">657</span> // Not going with further consistency check for tables when hbase:meta itself is not consistent.<a name="line.657"></a>
-<span class="sourceLineNo">658</span> LOG.info("Loading regionsinfo from the hbase:meta table");<a name="line.658"></a>
-<span class="sourceLineNo">659</span> boolean success = loadMetaEntries();<a name="line.659"></a>
-<span class="sourceLineNo">660</span> if (!success) return -1;<a name="line.660"></a>
-<span class="sourceLineNo">661</span><a name="line.661"></a>
-<span class="sourceLineNo">662</span> // Empty cells in hbase:meta?<a name="line.662"></a>
-<span class="sourceLineNo">663</span> reportEmptyMetaCells();<a name="line.663"></a>
-<span class="sourceLineNo">664</span><a name="line.664"></a>
-<span class="sourceLineNo">665</span> // Check if we have to cleanup empty REGIONINFO_QUALIFIER rows from hbase:meta<a name="line.665"></a>
-<span class="sourceLineNo">666</span> if (shouldFixEmptyMetaCells()) {<a name="line.666"></a>
-<span class="sourceLineNo">667</span> fixEmptyMetaCells();<a name="line.667"></a>
-<span class="sourceLineNo">668</span> }<a name="line.668"></a>
-<span class="sourceLineNo">669</span><a name="line.669"></a>
-<span class="sourceLineNo">670</span> // get a list of all tables that have not changed recently.<a name="line.670"></a>
-<span class="sourceLineNo">671</span> if (!checkMetaOnly) {<a name="line.671"></a>
-<span class="sourceLineNo">672</span> reportTablesInFlux();<a name="line.672"></a>
-<span class="sourceLineNo">673</span> }<a name="line.673"></a>
-<span class="sourceLineNo">674</span><a name="line.674"></a>
-<span class="sourceLineNo">675</span> // Get disabled tables states<a name="line.675"></a>
-<span class="sourceLineNo">676</span> loadTableStates();<a name="line.676"></a>
-<span class="sourceLineNo">677</span><a name="line.677"></a>
-<span class="sourceLineNo">678</span> // load regiondirs and regioninfos from HDFS<a name="line.678"></a>
-<span class="sourceLineNo">679</span> if (shouldCheckHdfs()) {<a name="line.679"></a>
-<span class="sourceLineNo">680</span> LOG.info("Loading region directories from HDFS");<a name="line.680"></a>
-<span class="sourceLineNo">681</span> loadHdfsRegionDirs();<a name="line.681"></a>
-<span class="sourceLineNo">682</span> LOG.info("Loading region information from HDFS");<a name="line.682"></a>
-<span class="sourceLineNo">683</span> loadHdfsRegionInfos();<a name="line.683"></a>
-<span class="sourceLineNo">684</span> }<a name="line.684"></a>
-<span class="sourceLineNo">685</span><a name="line.685"></a>
-<span class="sourceLineNo">686</span> // fix the orphan tables<a name="line.686"></a>
-<span class="sourceLineNo">687</span> fixOrphanTables();<a name="line.687"></a>
-<span class="sourceLineNo">688</span><a name="line.688"></a>
-<span class="sourceLineNo">689</span> LOG.info("Checking and fixing region consistency");<a name="line.689"></a>
-<span class="sourceLineNo">690</span> // Check and fix consistency<a name="line.690"></a>
-<span class="sourceLineNo">691</span> checkAndFixConsistency();<a name="line.691"></a>
+<span class="sourceLineNo">367</span> "hbase.hbck.lockfile.attempt.maxsleeptime", DEFAULT_LOCK_FILE_ATTEMPT_MAX_SLEEP_TIME));<a name="line.367"></a>
+<span class="sourceLineNo">368</span> createZNodeRetryCounterFactory = new RetryCounterFactory(<a name="line.368"></a>
+<span class="sourceLineNo">369</span> getConf().getInt("hbase.hbck.createznode.attempts", DEFAULT_MAX_CREATE_ZNODE_ATTEMPTS),<a name="line.369"></a>
+<span class="sourceLineNo">370</span> getConf().getInt(<a name="line.370"></a>
+<span class="sourceLineNo">371</span> "hbase.hbck.createznode.attempt.sleep.interval",<a name="line.371"></a>
+<span class="sourceLineNo">372</span> DEFAULT_CREATE_ZNODE_ATTEMPT_SLEEP_INTERVAL),<a name="line.372"></a>
+<span class="sourceLineNo">373</span> getConf().getInt(<a name="line.373"></a>
+<span class="sourceLineNo">374</span> "hbase.hbck.createznode.attempt.maxsleeptime",<a name="line.374"></a>
+<span class="sourceLineNo">375</span> DEFAULT_CREATE_ZNODE_ATTEMPT_MAX_SLEEP_TIME));<a name="line.375"></a>
+<span class="sourceLineNo">376</span> zkw = createZooKeeperWatcher();<a name="line.376"></a>
+<span class="sourceLineNo">377</span> }<a name="line.377"></a>
+<span class="sourceLineNo">378</span><a name="line.378"></a>
+<span class="sourceLineNo">379</span> private class FileLockCallable implements Callable<FSDataOutputStream> {<a name="line.379"></a>
+<span class="sourceLineNo">380</span> RetryCounter retryCounter;<a name="line.380"></a>
+<span class="sourceLineNo">381</span><a name="line.381"></a>
+<span class="sourceLineNo">382</span> public FileLockCallable(RetryCounter retryCounter) {<a name="line.382"></a>
+<span class="sourceLineNo">383</span> this.retryCounter = retryCounter;<a name="line.383"></a>
+<span class="sourceLineNo">384</span> }<a name="line.384"></a>
+<span class="sourceLineNo">385</span> @Override<a name="line.385"></a>
+<span class="sourceLineNo">386</span> public FSDataOutputStream call() throws IOException {<a name="line.386"></a>
+<span class="sourceLineNo">387</span> try {<a name="line.387"></a>
+<span class="sourceLineNo">388</span> FileSystem fs = FSUtils.getCurrentFileSystem(getConf());<a name="line.388"></a>
+<span class="sourceLineNo">389</span> FsPermission defaultPerms = FSUtils.getFilePermissions(fs, getConf(),<a name="line.389"></a>
+<span class="sourceLineNo">390</span> HConstants.DATA_FILE_UMASK_KEY);<a name="line.390"></a>
+<span class="sourceLineNo">391</span> Path tmpDir = new Path(FSUtils.getRootDir(getConf()), HConstants.HBASE_TEMP_DIRECTORY);<a name="line.391"></a>
+<span class="sourceLineNo">392</span> fs.mkdirs(tmpDir);<a name="line.392"></a>
+<span class="sourceLineNo">393</span> HBCK_LOCK_PATH = new Path(tmpDir, HBCK_LOCK_FILE);<a name="line.393"></a>
+<span class="sourceLineNo">394</span> final FSDataOutputStream out = createFileWithRetries(fs, HBCK_LOCK_PATH, defaultPerms);<a name="line.394"></a>
+<span class="sourceLineNo">395</span> out.writeBytes(InetAddress.getLocalHost().toString());<a name="line.395"></a>
+<span class="sourceLineNo">396</span> out.flush();<a name="line.396"></a>
+<span class="sourceLineNo">397</span> return out;<a name="line.397"></a>
+<span class="sourceLineNo">398</span> } catch(RemoteException e) {<a name="line.398"></a>
+<span class="sourceLineNo">399</span> if(AlreadyBeingCreatedException.class.getName().equals(e.getClassName())){<a name="line.399"></a>
+<span class="sourceLineNo">400</span> return null;<a name="line.400"></a>
+<span class="sourceLineNo">401</span> } else {<a name="line.401"></a>
+<span class="sourceLineNo">402</span> throw e;<a name="line.402"></a>
+<span class="sourceLineNo">403</span> }<a name="line.403"></a>
+<span class="sourceLineNo">404</span> }<a name="line.404"></a>
+<span class="sourceLineNo">405</span> }<a name="line.405"></a>
+<span class="sourceLineNo">406</span><a name="line.406"></a>
+<span class="sourceLineNo">407</span> private FSDataOutputStream createFileWithRetries(final FileSystem fs,<a name="line.407"></a>
+<span class="sourceLineNo">408</span> final Path hbckLockFilePath, final FsPermission defaultPerms)<a name="line.408"></a>
+<span class="sourceLineNo">409</span> throws IOException {<a name="line.409"></a>
+<span class="sourceLineNo">410</span><a name="line.410"></a>
+<span class="sourceLineNo">411</span> IOException exception = null;<a name="line.411"></a>
+<span class="sourceLineNo">412</span> do {<a name="line.412"></a>
+<span class="sourceLineNo">413</span> try {<a name="line.413"></a>
+<span class="sourceLineNo">414</span> return FSUtils.create(fs, hbckLockFilePath, defaultPerms, false);<a name="line.414"></a>
+<span class="sourceLineNo">415</span> } catch (IOException ioe) {<a name="line.415"></a>
+<span class="sourceLineNo">416</span> LOG.info("Failed to create lock file " + hbckLockFilePath.getName()<a name="line.416"></a>
+<span class="sourceLineNo">417</span> + ", try=" + (retryCounter.getAttemptTimes() + 1) + " of "<a name="line.417"></a>
+<span class="sourceLineNo">418</span> + retryCounter.getMaxAttempts());<a name="line.418">
<TRUNCATED>