0

we have redhat machines in cluster

redhat version is - 7.2

when we perform 

ps ax

on the machine we saw many kworker with state D

example

17165  ?  D  0:00  [kworker/u66:1]
2151  ?  D  0:58  [kworker/u33:0]
3866  ?  D  0:26  [kworker/u65:0]
14967  ?  D  0:00  [kworker/2:1]
2014  ?  D  0:10  [kworker/u65:1]
19021  ?  D  0:05  [kworker/u65:0]
26880  ?  D  9:43  [kworker/16:2]
7972  ?  D  9:04  [kworker/31:1]
12295  ?  D  2:01  [kworker/3:1]
13859  ?  D  0:11  [kworker/u65:3]

I must to say that we also face high load average on the CPU

based on that information , what should be the conclusion?

for example what we need to verify ? or to perform some tune for the OS?

shalom
  • 461
  • 13
  • 29
  • 2
    The first thing you should think when you see `D` in `ps` output is **disk**. – Michael Hampton Mar 27 '19 at 17:16
  • yes I am also thinking about this , but i see that iowait is normal , or maybe you have suggestions ? – shalom Mar 27 '19 at 17:23
  • 1
    We have no way to know if it is normal. We don't know what your workload is. – Michael Hampton Mar 27 '19 at 17:24
  • we saw also many on the machine we saw many kswapd0 with state D example 134 ? D 1099:58 [kswapd0] 134 ? D 1150:21 [kswapd0] 134 ? D 1099:59 [kswapd0] 134 ? D 1150:21 [kswapd0] 134 ? D 928:00 [kswapd0] 134 ? D 1099:59 [kswapd0] 134 ? D 1150:22 [kswapd0] – shalom Mar 27 '19 at 17:29
  • What is the user response time problem? Not the load average, but what are users complaining about that is slow. – John Mahowald Mar 27 '19 at 22:06

1 Answers1

0

If you have a kworker in d status and you are running other processes with i/o and the process is also in d status, chances are you can not kill either.

A reboot is required.

Michael G Waldron
  • 1