[hbase] branch branch-2 updated: HBASE-26945 Quotas causes too much load on meta for large clusters (#4576)

[bb...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

bbeaudreault pushed a commit to branch branch-2
in repository https://gitbox.apache.org/repos/asf/hbase.git


The following commit(s) were added to refs/heads/branch-2 by this push:
     new 409a7a44e8c HBASE-26945 Quotas causes too much load on meta for large clusters (#4576)
409a7a44e8c is described below

commit 409a7a44e8c2928930b5a04bfe86867e53732559
Author: Bryan Beaudreault <bb...@hubspot.com>
AuthorDate: Thu Jun 23 16:47:09 2022 -0400

    HBASE-26945 Quotas causes too much load on meta for large clusters (#4576)
    
    Signed-off-by: Xiaolin Ha <ha...@apache.org>
---
 .../org/apache/hadoop/hbase/quotas/QuotaCache.java | 28 +++++++++++++---------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/QuotaCache.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/QuotaCache.java
index 8bac7fd1a61..c8839ad61d1 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/QuotaCache.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/QuotaCache.java
@@ -27,12 +27,13 @@ import java.util.Map;
 import java.util.Set;
 import java.util.concurrent.ConcurrentHashMap;
 import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hbase.ClusterMetrics;
 import org.apache.hadoop.hbase.ClusterMetrics.Option;
-import org.apache.hadoop.hbase.MetaTableAccessor;
 import org.apache.hadoop.hbase.ScheduledChore;
 import org.apache.hadoop.hbase.Stoppable;
 import org.apache.hadoop.hbase.TableName;
 import org.apache.hadoop.hbase.client.Get;
+import org.apache.hadoop.hbase.client.RegionStatesCount;
 import org.apache.hadoop.hbase.regionserver.HRegionServer;
 import org.apache.hadoop.hbase.regionserver.RegionServerServices;
 import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
@@ -352,24 +353,29 @@ public class QuotaCache implements Stoppable {
      */
     private void updateQuotaFactors() {
       // Update machine quota factor
+      ClusterMetrics clusterMetrics;
       try {
-        int rsSize = rsServices.getConnection().getAdmin()
-          .getClusterMetrics(EnumSet.of(Option.SERVERS_NAME)).getServersName().size();
-        if (rsSize != 0) {
-          // TODO if use rs group, the cluster limit should be shared by the rs group
-          machineQuotaFactor = 1.0 / rsSize;
-        }
+        clusterMetrics = rsServices.getConnection().getAdmin()
+          .getClusterMetrics(EnumSet.of(Option.SERVERS_NAME, Option.TABLE_TO_REGIONS_COUNT));
       } catch (IOException e) {
-        LOG.warn("Get live region servers failed", e);
+        LOG.warn("Failed to get cluster metrics needed for updating quotas", e);
+        return;
+      }
+
+      int rsSize = clusterMetrics.getServersName().size();
+      if (rsSize != 0) {
+        // TODO if use rs group, the cluster limit should be shared by the rs group
+        machineQuotaFactor = 1.0 / rsSize;
       }
 
+      Map<TableName, RegionStatesCount> tableRegionStatesCount =
+        clusterMetrics.getTableRegionStatesCount();
+
       // Update table machine quota factors
       for (TableName tableName : tableQuotaCache.keySet()) {
         double factor = 1;
         try {
-          long regionSize =
-            MetaTableAccessor.getTableRegions(rsServices.getConnection(), tableName, true).stream()
-              .filter(regionInfo -> !regionInfo.isOffline()).count();
+          long regionSize = tableRegionStatesCount.get(tableName).getOpenRegions();
           if (regionSize == 0) {
             factor = 0;
           } else {