[
https://issues.apache.org/jira/browse/HBASE-2372?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12849449#action_12849449
]
Jean-Daniel Cryans commented on HBASE-2372:
-------------------------------------------
I'm not a fan of this loop:
{code}
+ while (!recovered) {
+ try {
+ FSDataOutputStream out = fs.append(p);
+ out.close();
+ recovered = true;
+ } catch (IOException e) {
+ LOG.info("Failed open for append, waiting on lease recovery: " + p, e);
+ try {
+ Thread.sleep(1000);
+ } catch (InterruptedException ex) {
+ // ignore it and try again
+ }
+ }
+ }
{code}
Looks like if there's any other kind of exception, like a missing block or
whatnot, we'll get stuck in that loop forever? I guess this also applies for
0.20
> hbase trunk should check dfs.support.append and use append() to recovery
> logfiles (as in branch)
> ------------------------------------------------------------------------------------------------
>
> Key: HBASE-2372
> URL: https://issues.apache.org/jira/browse/HBASE-2372
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: ryan rawson
> Assignee: ryan rawson
> Fix For: 0.21.0
>
> Attachments: HBASE-2372.txt
>
>
> this is a backport of the feature in the hlog recovery where we call
> fs.append(logfile); then fs.close() to recover a logfile in a situation where
> we run with HDFS-200 (ie: dfs.support.append = true)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
