[
https://issues.apache.org/jira/browse/CASSANDRA-15059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16813615#comment-16813615
]
Blake Eggleston commented on CASSANDRA-15059:
---------------------------------------------
I think both of these changes are worth considering, but I don’t think would be
appropriate to include them as part of this ticket. They are both out of scope
and too risky to be putting in 3.x, and I’d argue the same for 4.0 at this
point as well.
Gossiper is very brittle, critical to the operation of a cluster, and has
little to no test coverage. Any bugs that aren't caught by a manual review or
raise any red flags in the dtests will become production issues at some point.
> Gossiper#markAlive can race with Gossiper#markDead
> --------------------------------------------------
>
> Key: CASSANDRA-15059
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15059
> Project: Cassandra
> Issue Type: Bug
> Components: Cluster/Gossip
> Reporter: Blake Eggleston
> Assignee: Blake Eggleston
> Priority: Normal
>
> The Gossiper class is not threadsafe and assumes all state changes happen in
> a single thread (the gossip stage). Gossiper#convict, however, can be called
> from the GossipTasks thread. This creates a race where calls to
> Gossiper#markAlive and Gossiper#markDead can interleave, corrupting gossip
> state. Gossiper#assassinateEndpoint has a similar problem, being called from
> the mbean server thread.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
