Incredibly high load average on a Centos server

Question

I have a Centos 6.4 server that's being used as a Cassandra node in a production environment, it is still working and responding to queries etc, but the load average is rather "high" a.k.a. buggy? In any case, you have to see it for yourself:

incredibly high load average

Has anybody seen something like this before? Any clue on what's causing it?

Thanks in advance!

EDIT: Adding output of cat /proc/loadavg as requested:

[root@vdc-08 ~]# cat /proc/loadavg
2715599674950.05 1351413955738.05 471218463336.32 1/479 8382

@MikePurcell Indeed, although the server is still running and the numbers keep changing over time. I'm just waiting until somebody sheds some light on it before restarting the server. — Sergio Ayestarán, Feb 19 '15 at 00:13
@MatthewIfe Done Matthew, thanks for taking the time to see it — Sergio Ayestarán, Feb 19 '15 at 00:15

score 1 · Accepted Answer · edited Apr 13 '17 at 12:13

1

Almost certainly a kernel bug.

This question /proc/loadavg shows incorrect huge values suggested updating the kernel on the host and rebooting but did not specify what the root cause is.

edited Apr 13 '17 at 12:13

Community

1

answered Feb 19 '15 at 00:22

Matthew Ife

22,927
2
54
71

I'm not using cludlinux, this is a physical server running centos, but still the explanation is sound. – Sergio Ayestarán Feb 19 '15 at 00:30
Did it work? Updating the kernel (with subsequent reboot)? – Mike Purcell Feb 19 '15 at 01:09

score 0 · Answer 2 · answered Aug 16 '17 at 16:14

0

this was working for me:

add "elevator=noop" in kernel boot config.

High Load Average with modest CPU Utilization and almost no IO

answered Aug 16 '17 at 16:14

josy1024

1

2

"***was*** working" ... does that mean it does not anymore? – Pierre.Vriens Aug 16 '17 at 16:47

Incredibly high load average on a Centos server

2 Answers2