[
https://issues.apache.org/jira/browse/HBASE-2312?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13138002#comment-13138002
]
Phabricator commented on HBASE-2312:
------------------------------------
stack has requested changes to the revision "HBASE-2312 [jira] Possible data
loss when RS goes into GC pause while rolling HLog".
Looks good. A few minor items.
INLINE COMMENTS
src/main/java/org/apache/hadoop/hbase/HConstants.java:188 Should this define
be up here? Why not down in HLog? (I favor keeping defines in classes they
pertain to -- could be ugly though if we need to go cross packages to get at a
define. Master package would need to reach down into the wal package? If so,
I suppose, leave it up here in HConstants).
src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java:192 The
former code bracketed the return. Either put the return on same line as if or
else add back the brackets I'd say.
src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java:223 Why
not call server abort? Thats the usual idiom.
src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java:227 Why
this? And 30 seconds is a long time (Should it be configurable)?
src/main/java/org/apache/hadoop/hbase/master/MasterFileSystem.java:254 I
suppose this not the end of the world. A regionserver could add a new log I
suppose but chances of failed rename and RS adding new WAL are probably fairly
low.
src/main/java/org/apache/hadoop/hbase/regionserver/wal/SequenceFileLogWriter.java:126
Superfluous logging I'd say. Especially at info level. Log when we DON'T
have this feature (maybe do it debug; if every file open has one of these info
logs, then its going to generate lots of queries up in mailing lists, etc. At
DEBUG it won't seem as critical)
src/test/java/org/apache/hadoop/hbase/regionserver/wal/TestHLogSplit.java:831
Nice
REVISION DETAIL
https://reviews.facebook.net/D99
> Possible data loss when RS goes into GC pause while rolling HLog
> ----------------------------------------------------------------
>
> Key: HBASE-2312
> URL: https://issues.apache.org/jira/browse/HBASE-2312
> Project: HBase
> Issue Type: Bug
> Components: master, regionserver
> Affects Versions: 0.90.0
> Reporter: Karthik Ranganathan
> Assignee: Nicolas Spiegelberg
> Priority: Critical
> Fix For: 0.92.0
>
> Attachments: D99.1.patch
>
>
> There is a very corner case when bad things could happen(ie data loss):
> 1) RS #1 is going to roll its HLog - not yet created the new one, old one
> will get no more writes
> 2) RS #1 enters GC Pause of Death
> 3) Master lists HLog files of RS#1 that is has to split as RS#1 is dead,
> starts splitting
> 4) RS #1 wakes up, created the new HLog (previous one was rolled) and
> appends an edit - which is lost
> The following seems like a possible solution:
> 1) Master detects RS#1 is dead
> 2) The master renames the /hbase/.logs/<regionserver name> directory to
> something else (say /hbase/.logs/<regionserver name>-dead)
> 3) Add mkdir support (as opposed to mkdirs) to HDFS - so that a file
> create fails if the directory doesn't exist. Dhruba tells me this is very
> doable.
> 4) RS#1 comes back up and is not able create the new hlog. It restarts
> itself.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
