Server is going down because of jbd2/dm process consuming all I/O

[mlopez]

My web/email server is going down several times a day because of the process [jbd2/md1-8] which is using 99.99% IO according to iotop.

I've realized that MySQL (MariaDB 10.3.28) has something to do with it: the problem seems to be caused on database reading/writing.

I benchmarked the HD and it's ok, even better than others servers with no performance issues.
This server has just a few accounts (12 users, and a couple of them just use mail server, not mysql)

The specs are as follows:

CentOS Linux 7 (Core) x86_64
Kernel: 3.10.0-1160.15.2.el7.x86_64
CPU: Intel (Broadwell, IBRS) (4) @ 3.599GHz
Memory: 2439MiB / 14030MiB

How can I get rid of this io consumming process?

Thank you for your help.

 

[ZenHostingTravis]

Please see What is [jbd2/dm-3-8] and [kdmflush]? And why are they constantly on iotop? for more information aboug what jdb2 is.

In regards to your specific problem, the following link might assist you further: High server load - [jbd2/md1-8] using 99.99% IO

 

[cPRex]

Let us know if those links don't help!

 

[lakimens]

Sorry to bump and old thread. While this is informative and now I am aware that the flushing is causing the issue, how should it be resolved? I've read about some risky options such as disable journaling, or trade off resilience for speed in MySQL configuration, but what is the best fix for this? I am running it on an NVME Drive, and sometimes it's writing for hours at 70-90% IO usage.

 

[cPRex]

Hey there, @lakimens!!! Were you able to check that second link to get more details from the server to confirm what processes are causing the problem? This wouldn't be related to the cPanel tools on the system so my direct help is going to be a bit limited.