I am using solr 8.2 with zoo 3.4, and configured 5 node solr cloud with
around 100 collections each collection having ~20k documents.
These nodes are vm's with 6 core cpu and 2 cores per socket. All of sudden
seeing hikes on CPU's and which brought down some nodes (GONE state on solr
cloud and also faced latencies while trying to login to those nodes ssh)
Memory : 32GB and 20GB was allotted for jvm heap on solr config.
<filterCache class="solr.FastLRUCache"
size="50000"
initialSize="512"
autowarmCount="128"/>
<queryResultCache class="solr.LRUCache"
size="50000"
initialSize="512"
autowarmCount="128"/>
<documentCache class="solr.LRUCache"
size="50000"
initialSize="512"
autowarmCount="128"/>
<queryResultMaxDocsCached>200</queryResultMaxDocsCached>
<queryResultWindowSize>100</queryResultWindowSize>
<enableLazyFieldLoading>true</enableLazyFieldLoading>
<useColdSearcher>false</useColdSearcher>
<maxWarmingSearchers>4</maxWarmingSearchers>
These are just from the defaults that shipped with SOLR package.
One data point is that these nodes gets very frequent hits to them for
searching, so do I need to consider increasing the above sizes to get down
the CPU's and see more stable solr cloud?
--
Thanks & Regards,
Yaswanth Kumar Konathala.
yaswanth...@gmail.com
