On 6/5/2019 5:35 AM, vishal patel wrote:
We have 2 shards and 2 replicas in Live also have multiple collections. We are performing heavy search and update.
There is no information here about how many servers are serving those four shard replicas.
-> I have*attached*some query which takes time for executing. why does it take too much time? Due to the query length?
No attachments made it to the list. Attachments rarely make it -- you'll need to find some other way to share content.
-> Some times replica goes in recovery mode and from the log, we can not identify the issue but GC pause time 15 to 20 seconds. Ideally what should be GC pause time? GC pause time increase due to indexing or searching documents?
Individual GC pauses that are long are caused by having a large heap that undergoes a full collection. Long pauses from multiple collections are typically caused by a heap that's too small. When the heap is properly sized and GC is tuned well, full collections will be very rare, and the generation-specific collections will typically be very fast.
My Solr live data :
This indicates that your total size for shard1 is almost 400 gigabytes, and your total size for shard2 is almost 300 gigabytes.
If you have 400 or 700 GB of data on one server, then you will need a SIGNIFICANT amount of memory in that server, with most of it NOT allocated to the heap for Solr.
https://wiki.apache.org/solr/SolrPerformanceProblems#RAM Thanks, Shawn
