[
https://issues.apache.org/jira/browse/HBASE-7831?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13576916#comment-13576916
]
Sergey Shelukhin commented on HBASE-7831:
-----------------------------------------
What I had in mind is code like this:
{code}
if ((currentSeqNum != null)
&& (currentSeqNum.longValue() <= seqNumBeforeFlushStarts.longValue())) {
String errorStr = "Region " + Bytes.toString(encodedRegionName) +
"acquired edits out of order current memstore seq=" + currentSeqNum
+ ", previous oldest unflushed id=" + seqNumBeforeFlushStarts;
LOG.error(errorStr);
assert false : errorStr;
Runtime.getRuntime().halt(1);
}{code}
or
{code}
if (!checkFileSystem()) {
LOG.warn("Bad Filesystem, exiting");
Runtime.getRuntime().halt(1);
} {code}
> lightweight way to make RS commit suicide
> -----------------------------------------
>
> Key: HBASE-7831
> URL: https://issues.apache.org/jira/browse/HBASE-7831
> Project: HBase
> Issue Type: Improvement
> Reporter: Sergey Shelukhin
> Priority: Minor
>
> There are times in the life of many a region server when it realizes that it
> cannot go on any longer in the hostile, inhospitable environment. Or it might
> realize that there's something inherently wrong with it on the inside
> (probably because a developer screwed up). The only way out is killing itself.
> There's code in some places currently that does Runtime halt in such cases,
> and there may be code that calls HRegionServer method to do this.
> I think we need some easy-to-access (lightweight interface, or static) way to
> trigger reliable (no catching exceptions on the upper levels, etc.) RS death
> in such cases.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
