[
https://issues.apache.org/jira/browse/HBASE-2479?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12859951#action_12859951
]
Todd Lipcon commented on HBASE-2479:
------------------------------------
Here's my modified doWithRetries:
{code}
57 protected T doWithRetries()
58 throws IOException, RuntimeException {
59 List<IOException> exceptions = new ArrayList<IOException>();
60 for (int tries = 0; tries < this.master.numRetries; tries++) {
61 if (this.master.closed.get()) {
62 return null;
63 }
64 try {
65 // Seen an NPE here before, adding lots of logging to try to find
it!
66 if (this.master == null) {
67 ReflectionUtils.logThreadInfo(LOG, "master null", 1);
68 } else if (this.master.connection == null) {
69 ReflectionUtils.logThreadInfo(LOG, "connection null", 1);
70 } else if (m == null) {
71 ReflectionUtils.logThreadInfo(LOG, "m null", 1);
72 }
73
74 this.server =
75 this.master.connection.getHRegionConnection(m.getServer());
76 return this.call();
{code}
> NPE in master when meta is unavailable
> --------------------------------------
>
> Key: HBASE-2479
> URL: https://issues.apache.org/jira/browse/HBASE-2479
> Project: Hadoop HBase
> Issue Type: Bug
> Components: master
> Reporter: Todd Lipcon
> Attachments: npe-log.txt
>
>
> A couple times I've managed to get NPEs when processing region status
> messages due to META being unavailable. Will attach a log segment here.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
