1.3.4 - high cpu load - client login
would it be possible for you to try upgrading to Ubuntu 16.04, uninstalling all of the php5 packages and rerunning the fog installer, allowing the system to upgrade to php7? when i updated my Ubuntu 12.04 system to 16.04 with php7, my cpu load dropped to a quarter of what it was.
Wayne Workman last edited by
AH00161: server reached MaxRequestWorkers setting, consider raising the MaxRequestWorkers setting
That error I quoted, here’s some information on that:
That thread on that site suggests changing
MaxRequestWorkersinside that file, then restarting apache.
We probably have over 8,000 systems. I had the client checkin at 5 min and increased it to 10 min prior to posting. I appreciate the articles and I’m sure they’ll help. The memory of the server is fine and the swap is very small.
Wayne Workman last edited by Wayne Workman
Tom’s right. With that many hosts, you should have a FOG Client Checkin Time of like 5 minutes, if not longer. At my old job, we had about 2,000 systems tied to one big’ol fog setup, I set the check-in time to 15 minutes.
The setting is here:
Web Interface -> FOG Configuration -> FOG Settings -> FOG Client -> FOG_CLIENT_CHECKIN_TIME
You will also find this helpful:
And you might find this helpful too before you have the problem:
Then I might suggest you up the checkin time for the clients. I don’t know how many actual clients you have and I certainly have no clue when they are doing their cycles. The nice thing, however, is the checkin time is based on when the client starts up. Our default is 1 minute. This works fairly well, I think, up to 1000 clients (maybe more?)
You must take into account, however, one minute and a little extra. Our current setup gives a relatively random time of whatever the checkin time + 1 to 91 seconds.
So lets say your current time out is set to 60 seconds. The Maximum wait of a single host will only ever be 2 minutes and 31 seconds. The minimum wait of a single host will only ever be 61 seconds.
Given this information, you could have up to 151 hosts checking in every second. That would be 1 request per second.
While this isn’t horrible imagine 302 hosts? You could potentially have 2 hosts checking in every second and so on.
You can do the math from there of course, I’m just trying to help lay it out. I’d recommend upping the client checkin time to say 5 minute increments. (NOTE: This does not mean there is not a problem in the FOG Code, just what limited information I have I must start with what I believe is the most likely culprit.
Tom, we were at 1.3 RC 11 for a while. Almost all clients should be 11.5 or higher.
@ablohowiak What version of the fog client are these systems using?
The CPU Load issue you’re seeing would seem to be coming from an excess of queries to the FOG Server (leading me to think the high load issue is coming from the ‘legacy client’).
The new fog client should be much more responsive and usable while limiting the overall “work” the server needs to do.
The client login is my attempt at interpreting the apache error log.
We did briefly try 1.3.5 RC-1 with no improvement.
We have thousands of hosts.
How many hosts do you have in your network? This is useful to know, because the older reasoning of “MaxRequestWorkers” was due to having “persistent” sql connections (meaning apache was maintaining connections to mysql) which has since been removed (from 1.3.0 series to now).
What do you mean “client login”?
Have you tried the 1.3.5 RC’s to see if the problem still occurs?