All server web traffic lagging out when server load hits around 2.5 2.5 2.5

[maestroc]

I've had a cpanel server running on an eight core VPS for many years. I am migrating to a new server in a different data center due to server stability issues (data center keeps having "fires" and crap)...

The new server according to its specs is slightly more powerful than the old setup. 8 x Intel(R) Xeon(R) CPU E3-1270 V2 @ 3.50GHz and 8GB of memory.

The problem is that I have noticed on the new server that when the server load indicator spikes above 2.5 or 3 then suddenly I start getting notifications from my monitoring site that several of my websites are down. Sure enough, I check, and the site just sits there and sometimes takes several minutes to load, usually once the load has dropped.

This happened yesterday when I was using the transfer tool to move a large 15gb account over to the new server. Then it also happened today for no apparent reason that I could tell. Another time it apparently happened when it was running the overnight backups.

Both servers are running the same version of cpanel and Centos. Old server is running MariaDB 10.2, new one is running what was installed when the server was created, MySQL 5.7 I remember many years ago doing some tweaking to my.conf but have not done anything like that on the new server. Not sure if that is even a possible cause but thought I would mention it.

Any idea what things I should look at on the new server to fix this problem? Thank you in advance!

 

[kodeslogic]

When this happens again, you will have to identify the process using high resources using a top utility/command using the root user.

 

[cPRex]

What @kodeslogic said - the best way to get details on this will be to try and troubleshoot the issue while the load is high. It's nearly impossible to get reliable data after the issue has happened.

 

[maestroc]

I discovered the primary cause tonight as I caught the whole thing in the process of happening. One account is seeing a huge increase in traffic due to a very sensitive news article that had apparently gone viral. The thing that concerns me is that this spike in traffic to one account brought the whole server to its knees for several hours. Isn't apache/cpanel/etc. supposed to prevent high use by one account from bringing down everything else with it? I've never had this kind of traffic spike before so I'm unsure what I should have done differently or if something is somehow set incorrectly. I kept trying to kill all of the account's processes but they just came right back with more requests. I finally suspended the account just to let the server and the other accounts come back up.

 

[cPRex]

@maestroc - cPanel itself doesn't have any controls to prevent that. If the server has the resources, they can be used by one account.

What you're looking for is CloudLinux, which allows you to limit resources to individual domains. More details on that can be found here: Introduction