On 7/26/2019 7:16 AM, Rodrigo Oliveira wrote:
I have a cluster from Solr with Zookeeper (5 nodes - 48 Gb each node - Xms:28 Gb - Xmx:32 Gb). The bigger problem is my environment, because I am in a process of migration from MYSQL to SOLR.
Solr is not intended as a primary data store. There are things related to primary data store usage that MySQL can do which Solr either can't do at all or has a difficult time doing. Databases and search engines are each optimized for entirely different tasks.
It is reasonable to have your data in both a database and a search engine ... but to *switch* from a database to Solr sounds like a really bad idea.
Don't get me wrong... I'm one of the biggest fans of Solr you'll come across... but I am aware of its limitations as well as its strengths.
If the amount of data involved is small, using Solr as a primary data store might prove to be worthwhile ... but if I examine everything you have said, it doesn't sound like the amount of data is small.
Running OOM killer script for process 21252 for Solr on port 8983 Killed process 21252 I got it. Out Of Memory. Any workaround to this case? Any suggestion? Any tips? 10% of migration and Out Of Memory. Why the Solr it's consumed all memory?
This log is generated by the OOM killer script. It does not output any indication about WHY the error occurred. It simply indicates when the error occurred and what it did in response -- which is to terminate Solr.
There are several possible reasons for Java's OOME. Only a couple of those actually involve running out of memory. It might not be memory at all. But to find out, you will need to find the actual OutOfMemoryException in solr.log or one of the rotated versions (assuming it got logged at all), which will indicate the root issue.
There are precisely two solutions for OOME, and frequently only one of them is actually possible: Increase the resource that ran out, or figure out how to change the configuration so the program requires less of that resource. As already mentioned, you will need to figure out which resource was depleted.
If you can't find the actual exception, analyzing the GC log that Solr writes might help determine whether the depleted resource was heap memory.
Thanks, Shawn
