FTR, it looks like a "nodetool upgradesstables -a" address this issue.
It is still good to know that before running this command, any restart will hang for a long time. Hopping this will help someone, someday :). C*heers ! 2015-01-26 11:29 GMT+01:00 Alain RODRIGUEZ <arodr...@gmail.com>: > Hi guys, > > We migrate a cluster to 2.0.11 (From 1.2.18), in a rolling upgrade way. > > Now anytime I restart a node, it needs about 30' to start (350 GB > average). I used the debug level and see that we have a lot of "INDEX LOAD > TIME" lasting more than 200+ secs. It remembers me when I switched > index_interval to a new value and Cassandra had to rebuild indexes. > > Yet I did not change anything, and from what I read, index_interval per > table (2.0) is using the index_interval defined in cassandra.yaml (1.2). > Plus, it could have rebuild indexes the first time I restarted only, but I > experience this anytime I restart a node (any node, after multiple restart). > > Did any of you face this issue ? If not, any clue on what might be > happening ? > > I am trying to upgrade SSTable to see if it helps somehow. > > C*heers, > > Alain >
