High CPU usage state


#1

Topic says it...my linuxbox has entered this weird cpu state...where es is taking up a lot of CPU at idle. Logs are clean with no errors. Is there something I can look at to determine the cause? This is on a home machine with very little traffic. Thank you.


(Mark Walkom) #2

There's a few threads on the same topic that'd be worth checking.

But; hot threads, GC. How are you monitoring it?


#3

HI Mark....you're a good man for always jumping into my lame threads. I'm using htop and gkrellm. A stop/start of the service has cured this for now...I only have data from June 11th, so I don't think it's a capacity issue. Again, nothing shows in the logs...just...suddenly cpu pegs out...this is on ES 1.4.5...going to upgrade soon. Thanks again mark.


(Mark Walkom) #4

Like I mentioned, hot threads is a good place to start as well.


#5

Thanks Mark....after researching, now I know what hot threads are :smile: When this happens again I will:

links http://127.0.0.1:9200/_nodes/hot_threads?pretty

and report the findings here. Thanks again Mark...you always help me out.


#6

Here's what I got:

0.7% (3.5ms out of 500ms) cpu usage by thread 'elasticsearch[Ringleader][scheduler][T#1]'
 10/10 snapshots sharing following 9 elements


0.0% (198.9micros out of 500ms) cpu usage by thread 'elasticsearch[Ringleader][transport_client_worker][T#2]{New I/O worker #2}'
 10/10 snapshots sharing following 15 elements


0.0% (172.8micros out of 500ms) cpu usage by thread 'elasticsearch[Ringleader][transport_client_timer][T#1]{Hashed wheel timer #1}'
 10/10 snapshots sharing following 5 elements

::: [logstash-gateway-28922-4324][fjA2elXlSc66pN9K2-LpTA][gateway][inet[/192.168.1.253:9301]]{data=false, client=true}
Hot threads at 2015-08-05T10:50:48.523Z, interval=500ms, busiestThreads=3, ignoreIdleThreads=false:

0.1% (593.1micros out of 500ms) cpu usage by thread 'elasticsearch[logstash-gateway-28922-4324][[http_server_worker.default]][T#2]{New I/O worker #7}'
 10/10 snapshots sharing following 15 elements


0.0% (211.6micros out of 500ms) cpu usage by thread 'Ruby-0-Thread-9: /opt/logstash/vendor/bundle/jruby/1.9/gems/stud-0.0.17/lib/stud/buffer.rb:92'
 10/10 snapshots sharing following 31 elements


0.0% (180.4micros out of 500ms) cpu usage by thread 'elasticsearch[logstash-gateway-28922-4324][transport_client_timer][T#1]{Hashed wheel timer #1}'
 10/10 snapshots sharing following 5 elements

(Andrei Stefan) #7

@DigiAngel those hot threads are taken while the CPU usage is high? From those hot threads it doesn't look like the CPU is busy.


#8

Yea that's what I thought too....I've bumped the service.....next time it happens I'll dump out ps aux as well. Thank you.


#9

Here's what top has to say:

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                                                                             
               
24197 elastic+  20   0  917144 555264   5256 S  52.1 18.0 237:49.77 java                                                                                                                                                                                  

using 50% cpu

and from elasticsearch-parabedic:

 Dazzler
    
  ID: SUS_sfKsRcm7vNZ5ERFJxQ
  IP: inet[/127.0.0.1:9200]
  Host: 
  Load: 3.850
  Size: 
  Docs: 5 031 059
  Heap: 
  /

Thank you.


#10

Here's what top has to say:

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                                                                             
               
24197 elastic+  20   0  917144 555264   5256 S  52.1 18.0 237:49.77 java                                                                                                                                                                                  

using 50% cpu

and from elasticsearch-parabedic:

 Dazzler
    
  ID: SUS_sfKsRcm7vNZ5ERFJxQ
  IP: inet[/127.0.0.1:9200]
  Host: 
  Load: 3.850
  Size: 
  Docs: 5 031 059
  Heap: 
  /

Thank you.


(Mark Walkom) #11

52% isn't high.

What does hot threads say?


#12

As soon as I removed all data besides the last 10 days this immediately cleared up and I'm back to normal cpu usage. Hot threads had nothing to say....same as above...almost nothing so...eh...guess I'll upgrade to ES 1.7.1 and see what happens. Thanks Mark and all!


(system) #13