You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@accumulo.apache.org by GitBox <gi...@apache.org> on 2019/12/03 14:53:06 UTC

[GitHub] [accumulo] jzgithub1 commented on issue #1443: Fixes #1220 - Removes ZooKeeper watches on table configs when table is deleted.

jzgithub1 commented on issue #1443: Fixes #1220  -  Removes ZooKeeper watches on table configs when table is deleted.
URL: https://github.com/apache/accumulo/pull/1443#issuecomment-561202650
 
 
   > The pull request does not appear to be doing what it purports to do in the comments/log message/subject. That is, it does not appear to remove watches... but does appear to avoid adding some possibly unnecessary ones on nodes that don't yet exist. I _think_ that works because the case of watching for new child node creation is covered by the fact that ZooCache clears out cached child nodes when the parent node watcher is triggered, so the parent node watcher is sufficient.
   > 
   
   @ctubbsii ,  I have verified with my Uno instance and the ZooKeeper four letter command 'wchp' that this pull request does indeed remove all of the ZooKeeper watches related to an individual table when that table is deleted.  I verified with the 'createtable' and 'deletetable' commands in the accumulo shell.  There is nothing academic about it.  If 'wchp' is giving me an accurate state of the watchers in ZooKeeper then this pull request is removing watches and associated session information when tables are deleted.  Our current code just leaves all the table configuration watches for a table to remain watched all of the time when a table is deleted even though the actual Zoonodes are gone.   I can demo this PR at our next Hackathon.  PR #1422  which implements your idea in #1225  works well too and eliminates all the watches for sure.   If someone considers that never removing the watches when tables are deleted to be the central problem of running out of memory in Zookeeper,  than we can solve the issue of crashing ZooKeepers with this PR.  That was the hypothesis  that @keith-turner put forth in #1220 .  
   
   
   

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services