From 4ebbfa3d60c86fa7cac41ad582ab1b6edc990582 Mon Sep 17 00:00:00 2001 From: Peter Somogyi Date: Thu, 29 Nov 2018 18:04:53 +0100 Subject: [PATCH] HBASE-21518 TestMasterFailoverWithProcedures is flaky Signed-off-by: Sean Busbey --- .../hadoop/hbase/util/JVMClusterUtil.java | 24 +++++++++++++------ 1 file changed, 17 insertions(+), 7 deletions(-) diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/util/JVMClusterUtil.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/util/JVMClusterUtil.java index 8c92f6660c0..7518d65a198 100644 --- a/hbase-server/src/main/java/org/apache/hadoop/hbase/util/JVMClusterUtil.java +++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/util/JVMClusterUtil.java @@ -249,14 +249,24 @@ public class JVMClusterUtil { // Do backups first. JVMClusterUtil.MasterThread activeMaster = null; for (JVMClusterUtil.MasterThread t : masters) { - if (!t.master.isActiveMaster()) { - try { - t.master.stopMaster(); - } catch (IOException e) { - LOG.error("Exception occurred while stopping master", e); + // Master was killed but could be still considered as active. Check first if it is stopped. + if (!t.master.isStopped()) { + if (!t.master.isActiveMaster()) { + try { + t.master.stopMaster(); + } catch (IOException e) { + LOG.error("Exception occurred while stopping master", e); + } + LOG.info("Stopped backup Master {} is stopped: {}", + t.master.hashCode(), t.master.isStopped()); + } else { + if (activeMaster != null) { + LOG.warn("Found more than 1 active master, hash {}", activeMaster.master.hashCode()); + } + activeMaster = t; + LOG.debug("Found active master hash={}, stopped={}", + t.master.hashCode(), t.master.isStopped()); } - } else { - activeMaster = t; } } // Do active after.