Do full gc frequently?


(famoss) #1

Today,i find my es node do full gc frequently,but there are no other logs. I want to use jmap to dump the memory,but the dump file's size is 19GB,mat can't analysis the file.

I use _nodes/stats to get the node info,but fielddata and query cache just use 2GB size.

How can i to get the full gc's reason?

Here is the es log.

[2016-11-24 17:06:48,451][INFO ][monitor.jvm              ] [es140.213-2] [gc][old][6221773][2589] duration [18.7s], collections [2]/[18.9s], total [18.7s]/[17.3m], memory [22.9gb]->[11.4gb]/[24.8gb], all_pools {[young] [119.7mb]->[96.2mb]/[1.4gb]}{[survivor] [191.3mb]->[0b]/[191.3mb]}{[old] [22.6gb]->[11.3gb]/[23.1gb]}
[2016-11-24 17:09:18,663][WARN ][monitor.jvm              ] [es140.213-2] [gc][old][6221898][2594] duration [21.2s], collections [2]/[21.3s], total [21.2s]/[17.6m], memory [23.9gb]->[11.7gb]/[24.8gb], all_pools {[young] [756.3mb]->[222.3mb]/[1.4gb]}{[survivor] [191.3mb]->[0b]/[191.3mb]}{[old] [23gb]->[11.5gb]/[23.1gb]}
[2016-11-24 17:11:11,928][WARN ][monitor.jvm              ] [es140.213-2] [gc][old][6221990][2597] duration [18.7s], collections [1]/[19.5s], total [18.7s]/[17.9m], memory [22.8gb]->[11.4gb]/[24.8gb], all_pools {[young] [95.6mb]->[68.7mb]/[1.4gb]}{[survivor] [191.3mb]->[0b]/[191.3mb]}{[old] [22.5gb]->[11.3gb]/[23.1gb]}
[2016-11-24 17:13:35,252][WARN ][monitor.jvm              ] [es140.213-2] [gc][old][6222112][2601] duration [17.5s], collections [1]/[17.7s], total [17.5s]/[18.2m], memory [23.5gb]->[11.2gb]/[24.8gb], all_pools {[young] [459.8mb]->[303.8mb]/[1.4gb]}{[survivor] [191.3mb]->[0b]/[191.3mb]}{[old] [22.9gb]->[10.9gb]/[23.1gb]}
[2016-11-24 17:15:35,798][WARN ][monitor.jvm              ] [es140.213-2] [gc][old][6222213][2604] duration [17.2s], collections [1]/[17.6s], total [17.2s]/[18.5m], memory [23.7gb]->[11.1gb]/[24.8gb], all_pools {[young] [938.4mb]->[124.5mb]/[1.4gb]}{[survivor] [191.3mb]->[0b]/[191.3mb]}{[old] [22.6gb]->[11gb]/[23.1gb]}

(Adrien Grand) #2

Unfortunately this is one of the hardest questions to answer. Maybe other tools than mat would support analyzing such a large dump?


(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.