phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <>
Subject [jira] [Commented] (PHOENIX-3947) Increase scan time out for partial index rebuild and retry only once
Date Fri, 07 Jul 2017 17:14:00 GMT


James Taylor commented on PHOENIX-3947:

Thanks for the patch, [~samarthjain]. Here's some feedback:
- These lines should not have been removed (as otherwise we'd ignore the potential override
of these settings):
-        enableRebuildIndex = env.getConfiguration().getBoolean(QueryServices.INDEX_FAILURE_HANDLING_REBUILD_ATTRIB,
-            QueryServicesOptions.DEFAULT_INDEX_FAILURE_HANDLING_REBUILD);
-        rebuildIndexTimeInterval = env.getConfiguration().getLong(QueryServices.INDEX_FAILURE_HANDLING_REBUILD_INTERVAL_ATTRIB,
- Add a lower level unit test to PropertiesUtilTest for PropertiesUtil.extractProperties()
- IMHO, it's somewhat dangerous/unexpected that PropertiesUtil.extractProperties() modifies
{{props}} in-place. I think it'd be better if it made a new Properties instance by using PropertiesUtil.deepCopy(props)
and returns the new properties. Make sure all occurrences assign the return value (most do,
but your particular usage doesn't).
- Upon failure of an index to rebuild, we should disable the index and clear the INDEX_DISABLE_TIMESTAMP
to prevent further retries. You can use IndexUtil.setIndexDisableTimeStamp() for this using
0 for the minTimeStamp and PIndexState.DISABLE as the newState.

> Increase scan time out for partial index rebuild and retry only once
> --------------------------------------------------------------------
>                 Key: PHOENIX-3947
>                 URL:
>             Project: Phoenix
>          Issue Type: Bug
>            Reporter: James Taylor
>            Assignee: Samarth Jain
>              Labels: globalMutableSecondaryIndex
>         Attachments: PHOENIX-3947.patch, PHOENIX-3947_v2.patch
> The scan done from MetaDataRegionObserver needs to have a higher timeout so that it can
always complete successfully. The retries should be limited to one too. Upon failure, we should
disable the index and set the INDEX_DISABLE_TIMESTAMP to zero to prevent further attempts
to rebuild the index (as a failure would be an indication that something is awry and manual
intervention should be required to rebuild the index completely).

This message was sent by Atlassian JIRA

View raw message