VPS Performance

[griffeymania]

I have been a KnownHost VPS customer for over a year now and have generally been pleased with the service; however, I'm running into a server performance issue and need some advice.

I currently host around 12 websites, with a total of 63 mailboxes for these domains. I have 512 MB of dedicated RAM and have never seen the amount of free RAM dip below 250 MB when running top. Recently, however, I've been experiencing periods of time where the load average spikes and my Idle Wait times run at about 25-30%. The system slows to a crawl and becomes nearly unresponsive... but usually resolves itself within 15 minutes.

I submitted a ticket to KnowHost tech support and they requested that I do a [FONT=Verdana, Arial, Helvetica] ps auxxxfx and [/FONT][FONT=Verdana, Arial, Helvetica] pstree -aup when the issue occurred. They concluded that the issue was related to users popping their mail off the server. Here is a quote:

[/FONT][FONT=Verdana, Arial, Helvetica]"It seems that the many mails are fetched from the POP server and because of that the resource usage of mail server increases.

We can see many fail count for the VPS resource parameter privvmpages, which controls the memory usage.

--------------------------------
root@ns1 [~]# cat /proc/user_beancounters
Version: 2.5
uid resource held maxheld barrier limit failcnt
1269: kmemsize 7332167 7577717 29614080 30474240 1364981
lockedpages 0 0 384 384 0
privvmpages 91876 97870 262144 266076 3172272
--------------------------------[/FONT]"

What I don't understand is why my available memory in top always appears to be around 250 MB free but the VPS is running into its limit. Tech support says to increase my RAM or remove mail accounts, but won't tell me how much RAM I really need. I would appreciate *any* opinions on what would best help resolve my server performance.

Thanks,
Brady

 

[KH-Paul]

Hi Brady,

Could you please PM me your ticket(s) number so I can get familiar with the situation?

The output you've provided shows fail counters for kmemsize and privvmpages resources. Kmemsize controls amount of low (or kernel) memory available for processes running inside the VPS and in most cases such alerts can be registered in case if you have busy web sites with url_include pointing to local sites and/or if there is extensive use of MySQL databases with persistent connections enabled. These are two common reasons why failure can be seen but there could be others.
As for privvmpages - this parameter limits amount of memory that can be allocated and potentially used by processes running on the system. Failures will be reported if processes running inside the VPS are trying to allocate more memory that hosting plan allows.

Please either PM me your ticket number(s) or update the ticket asking it to be escalated to Paul so I can research this a bit more in attempt to find out what might be happening there.

 

[jamesp57]

I would like to know the outcome of this as well. We have been having the same problem for a week or two. I have been in touch with tech support, but they were not sure what was causing the problem.

It's very frustrating.

 

[KH-Paul]

Hi James,

Please PM me your ticket number, I'll take a look at what is going on there.

Regards,
Paul