On 9/14/23 07:16, rajani m wrote:
Karl, I like your suggestion and will keep in mind for k8 cluster
implementation. For now, interestingly, shards remain in recovery state
until the first searcher
<https://solr.apache.org/guide/7_5/query-settings-in-solrconfig.html#query-related-listeners>
listener
completes replaying queries and which avoids zk from sending any traffic to
it, putting production traffic queries there to help warmup caches before
zk gets it as active shard solves it. Thank you all for the suggestions.
If you are encountering long transaction log replay times, you can solve
that with this config that you can find under updateHandler in
solrconfig.xml in the example configs:
<autoCommit>
<maxTime>15000</maxTime>
<openSearcher>false</openSearcher>
</autoCommit>
I like to raise that to 60000 just so there is less load on the system
during heavy indexing, but 15 seconds also works because that kind of
commit happens VERY quickly.
Thanks,
Shawn
