[GitHub] [accumulo] EdColeman commented on a change in pull request #1158: Require x tservers on start

[GitBox <gi...@apache.org>]

EdColeman commented on a change in pull request #1158: Require x tservers on start
URL: https://github.com/apache/accumulo/pull/1158#discussion_r284017180
 
 

 ##########
 File path: server/master/src/main/java/org/apache/accumulo/master/Master.java
 ##########
 @@ -1436,6 +1441,86 @@ public void run() {
     log.info("exiting");
   }
 
+  /**
+   * Allows property configuration to block master start-up waiting for a minimum number of tservers
+   * to register in zookeeper. It also accepts a maximum time to wait - if the time expires, the
+   * start-up will continue with any tservers available. The following properties are used to
+   * control the behaviour:
+   * <ul>
+   * <li>MASTER_STARTUP_TSERVER_AVAIL_MIN_COUNT - when set to 0, no blocking occurs (default
+   * behaviour) otherwise will block until the number of tservers are available.</li>
+   * <li>MASTER_STARTUP_TSERVER_AVAIL_MAX_WAIT - time to wait in milliseconds. When set to 0, will
+   * block indefinitely.</li>
+   * </ul>
+   *
+   * @throws InterruptedException
+   *           if interrupted while blocking, propagated for caller to handle.
+   */
+  private void blockForTservers() throws InterruptedException {
+
+    long waitStart = System.currentTimeMillis();
+
+    AccumuloConfiguration accConfig = serverConfig.getConfiguration();
+    long minTserverCount = accConfig.getCount(Property.MASTER_STARTUP_TSERVER_AVAIL_MIN_COUNT);
+
+    if (minTserverCount <= 0) {
+      log.info(
+          "tserver availability check disabled, contining with-{} servers." + "To enable, set {}",
+          tserverSet.size(), Property.MASTER_STARTUP_TSERVER_AVAIL_MIN_COUNT.getKey());
+      return;
+    }
+
+    long maxWait = accConfig.getTimeInMillis(Property.MASTER_STARTUP_TSERVER_AVAIL_MAX_WAIT);
+
+    if (maxWait <= 0) {
+      log.info("tserver availability check set to block indefinitely, To change, set {} > 0.",
+          Property.MASTER_STARTUP_TSERVER_AVAIL_MAX_WAIT.getKey());
+      maxWait = Long.MAX_VALUE;
+    }
+
+    // honor Retry condition that initial wait < max wait, otherwise use small value to allow thread
+    // yield to happen
+    long initialWait = Math.min(50, maxWait / 2);
+
+    Retry tserverRetry =
+        Retry.builder().infiniteRetries().retryAfter(initialWait, TimeUnit.MILLISECONDS)
+            .incrementBy(15_000, TimeUnit.MILLISECONDS).maxWait(maxWait, TimeUnit.MILLISECONDS)
+            .logInterval(30_000, TimeUnit.MILLISECONDS).createRetry();
+
+    log.info("Checking for tserver availability - need to reach {} servers. Have {}",
+        minTserverCount, tserverSet.size());
+
+    boolean needTservers = tserverSet.size() < minTserverCount;
+
+    while (needTservers && tserverRetry.canRetry()) {
+
+      tserverRetry.waitForNextAttempt();
+
+      needTservers = tserverSet.size() < minTserverCount;
+
+      log.info(
 
 Review comment:
   This was intentional.  The intent was that on getting lock there would be a record in the log of the number of servers available and what the min threshold was set to (even if disabled with 0) - so if there were future questions / issues, what the initial state would be captured.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services