[jira] [Commented] (SOLR-13814) Rebalance Ram usages for Solr Cloud

["Shawn Heisey (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-13814?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16943460#comment-16943460 ] 

Shawn Heisey commented on SOLR-13814:
-------------------------------------

It is completely normal for any Java program to eventually use all of the max heap that you have told it that it can have.  That is the nature of Java.  If you want it to use less memory, give it a lower max heap, and if it's possible for the program to run with less heao then it will.

If you're looking at the physical memory usage, if your indexes are big enough, it is also completely normal for Solr to cause the OS to use all of the un-allocated physical memory for disk caching purposes.  With terabytes of index, your indexes definitely qualify as big enough.  If any program on your system suddenly needs more memory than is currently available, the OS will sacrifice the buffers/cache memory to allow the program to use it.


> Rebalance Ram usages for Solr Cloud
> -----------------------------------
>
>                 Key: SOLR-13814
>                 URL: https://issues.apache.org/jira/browse/SOLR-13814
>             Project: Solr
>          Issue Type: Improvement
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: SolrCloud
>    Affects Versions: 6.6
>            Reporter: Hoan Tran Van
>            Priority: Minor
>
> I have a SolrCloud with 9 nodes on 9 physical servers. the server 2, 4, 5 usually have high RAM usage > 92%. I restart but they will increase after 1-3 days, others servers never reach 90% RAM. 9 severs are similar. 9 solr nodes have same solr.in.sh configuration. 9 solr nodes use DIH for importing from Oracle. I compare some metrics but there are not much different (except server 1,7 have less core than others). How to rebalance ram usage for 9 servers?
> |Server|Total request last 12h|Total update last 12h|Replica|Leader|Total core|index size (TB)|Number of doc (billion)|-/+ buffers/cache (GB)|Ram free (MB)|Total RAM|Cache size|
> |Server 1|12M|10M|17|37|54|2.3|190|35|1288|252|0|
> |Server 2|24M|92M|57|18|75|3.2|270|18|655|252|0|
> |Server 3|23M|95M|38|34|72|2.98|260|53|2062|252|0|
> |Server 4|16M|99M|28|48|76|3.03|270|30|1329|252|0|
> |Server 5|15M|15M|67|3|70|2.87|255|15|1535|252|0|
> |Server 6|15M|11M|37|32|69|2.85|250|71|828|252|0|
> |Server 7|14M|10M|10|61|71|2.3|191|38|2459|252|0|
> |Server 8|17M|12M|61|12|73|3|266|86|1648|252|0|
> |Server 9|15M|12M|16|59|75|3.11|270|58|778|252|0|



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org