Jon Haddad created CASSANDRA-15312:
--------------------------------------
Summary: changing key cache size can break a node after restart
Key: CASSANDRA-15312
URL: https://issues.apache.org/jira/browse/CASSANDRA-15312
Project: Cassandra
Issue Type: Bug
Reporter: Jon Haddad
I observed this on a cluster we were doing a rolling restart on. I had
intended on digging further into this, but haven't had the time, and wanted to
document it here in case anyone saw similar behavior. Here's what was done:
# We resized the key cache using nodetool, and updated the config without
restarting. I believe the same settings in the config were applied via
nodetool, but I'm not 100% sure.
# Several days went by, so the cache was flushed.
# The node was restarted.
What we observed is that when the node was brought back up, it didn't know
about any of its tokens. Restarting the node didn't help.
The problem went away after we stopped the node, removed the saved key cache
data, and brought the node back up.
I'm not sure if it's necessary to change the key cache size in the yaml to
cause this issue, or to change it via nodetool, or both.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
