Server instability seems to be related to CSF or LFD

[davidp]

Good morning, I manage a couple of whm servers on centos 7 with imunify360 and csf, within the last couple of days the servers have become really unstable. To the best of my ability it seems like I have isolated the cause to be is LFD causing the servers to run out of memory and the way was able to prove it is the following.

A couple of days ago I was receiving e-mails from one of my servers that lfd failed to start, so I went into service manager and unchecked lfd to see what would happen and the server was fine for almost 24 hours, because last night I restarted csf due to it being in an error state and must have not caught that lfd had started backup and the server had become unstable at around 2:30am until it was rebooted by the provider.

This time I made sure lfd is off on both and will monitor, but the odd part is that this started happening I want to say Friday or Saturday of last week, but only really got unstable this week, from what I can tell all the hardware is fine as there are no weird post messages or anything like that and the servers run fine for a little bit (roughly a couple of hours) it depends, and then all of sudden server load starts rapidly peaking and on the console I get nothing but trash, sometimes I get firewall stuff spewing across the console.

Sometimes I am able to issue a three finger salute to soft reboot it, but a majority of the time I have to do a hard reset, because it becomes so unresponsive even on the console.

Has anyone heard or experienced this recently? No weird configurations that I am aware of or config changes that were done prior to the issues occurring.

[davidp]

Please note that I have seen a server degrade in as little as 30 mins, with average loads in the 700s or better.

[alexf]

In my experience with a WHM server on CentOS7, sudden issues usually revolve around a failed or hung upcp script. I would chexk the upcp log files dating back to when this started. You can find the log file at /var/cpanel/ updatelogs/.

It can’t hurt to start with this troubleshooting step.