> On Feb 1, 2019, at 1:51 AM, Troels Arvin <tro...@arvin.dk> wrote:
>
> Hello,
>
> I think I understand why one needs to regularly run "nodetool repair" on
> normal Cassandra installations with more than one node.
For people following along at home, the reason is to make sure tombstones make
it to all hosts before gc grace seconds expires
>
> But am I right about the following?
> In single-node Cassandra installations, it is irrelevant to run "nodetool
> repair" cron jobs.
It shouldnt be needed, because the tombstones must already be there.
One weird edge case note though: there is a setting on some versions of
Cassandra that only allows Cassandra to purge tombstones that have been
repaired. If you set that on a single node cluster, it’s impossible to satisfy
that condition, since you can’t repair one host, so you can’t purge any of
those tombstones. It’s effectively like setting infinite gc grace. If you’ve
turned that setting on for a single host cluster, be sure to turn it off.
---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org
For additional commands, e-mail: user-h...@cassandra.apache.org
