[
https://issues.apache.org/jira/browse/HBASE-2363?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12848392#action_12848392
]
Todd Lipcon commented on HBASE-2363:
------------------------------------
Another example is in HLog where we swallow exceptions writing the split logs.
Before fixing HBASE-2355 I was seeing a case where exceptions closing the
output of log splitting were ignored and then data was silently lost (even with
sync). I would rather have had the HMaster shut itself down.
> Add configuration for tunable dataloss sensitivity
> --------------------------------------------------
>
> Key: HBASE-2363
> URL: https://issues.apache.org/jira/browse/HBASE-2363
> Project: Hadoop HBase
> Issue Type: Improvement
> Affects Versions: 0.20.3, 0.21.0
> Reporter: Todd Lipcon
>
> There are many cases in HBase when the process detects that some data is "not
> quite right". We often have two choices: (a) throw up our hands and exit, or
> (b) log a warning and push through the issue, potentially losing data. I
> think some users are very sensitive to dataloss and would prefer that the
> system become unavailable rather than continue with potentially lost data.
> Other users just want the system to stay up, and if they lose a log segment
> it's fine.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
