Details
Description
Sometimes an index/view gets corrupted and the error
{"error":"error","reason":"inconsistent_state"}gets returned from query. Seems to happen under moderate load conditions. And often!. ns-diag report was huge, so I attach errors.1 file and log.8 files, which seem to contain the most recent logs and a bunch of errors around the time the problem occurred, 2012-04-10 22:22:26
At the time a FLUSH had just been performed, then a moderate load of data was applied then and a query to access a view was issued, which would have caused an index to kick off
The only way I have been able to resolve the problem is to flush and then restart each node.
PS It would be nice to be able to set an index as continuous and have it incrementally build say every 0.5 or 1 sec, this would prevent users from having to issues queries to perform that task
Thanks for the great work
Attachments
Issue Links
- duplicates
-
MB-4717 flush_all doesn't work on the latest 2.0 builds
- Closed