High Server Load?

[uberamd]

I have been tracking server load for the last nearly 24-hours on my hostgator shared server (I realize it isn't long but still, it does show trends) and here is what I came up with (note this image updates dynamically):


I am curious as to what the server load should be. Sources I have read state that a good server has a load of 1.0 or less, but I am guessing this does not apply to HG shared servers. So, if a HG tech could weigh in, what is the ideal load you are looking to have on your servers? Various dedicated servers I manage as well as other shared hosts I use have much, much lower loads than my HG server.

Here is a graph of one of my dedicated servers, and the spike was caused by load testing (2500 requests):


Just curious.

[slapshotw]

Server loads are not relative and not percentages. This means that a 2, for instance, can be handled really really well by a new server, but may crash an older server. Because of this, unless the hardware specs of the servers you're comparing are exactly the same, you can't compare the load for one against the load for another.

HG servers are very, very powerful. They can easily handle loads of 6 or 8 without problem.

[GatorPatrick]

If I was on shift and working I wouldn't even look twice at a box with the loads you've got there. IMO you're right where you should be. Our servers really won't even bat an eye until the load is up around the 20 range.

[calum]

HG servers seem to preform fine up to around 10, or even 15.

But if it gets too high then tell staff, mine once had a run-away script or something and it got to 30, I told HG and it got to 45 and ran at 45 for 10 minutes.

It was slow to load sites etc. and some things didn't work.

[GatorDaveC]

We have a monitor for all of the servers, we try to make sure that all of the servers have a load lower then 10.00. Anything that is 10 or higher will start flashing on our screens, as well as if a service is down.

If there is a run away PHP script that uses over 150MB of ram or uses over 30 seconds of CPU time. The script will automatically be killed.

Now CPU time isn't done in regular seconds, kind of put it this way... If you use 50% of the CPU you would be able to use that process for a full minute. Most scripts never use 100% CPU usage since there are other processes running on the server.

If your PHP script is running at 0% CPU usage, and hanging its most likely that you have something being fetched that isn't connecting, or an SQL issue with your query.

Nice graph btw.

[IronWarrior]

ooo How are you guys getting the server loads like this from?

[GatorDaveC]

Originally Posted by IronWarrior

ooo How are you guys getting the server loads like this from?

There are many ways. Here are a couple of ways.

-- The uptime command --
gateway04 plugins # uptime
04:18:45 up 39 days, 1:33, 1 user, load average: 0.80, 1.26, 1.83
gateway04 plugins #
---

-- /proc/loadavg --
gateway04 plugins # cat /proc/loadavg
0.72 1.19 1.78 1/122 25212
gateway04 plugins #
---

You could use the system command for "uptime", or a more efficient way for PHP would be to read "/proc/loadavg" in PHP.

[calum]

If you are on shared you can also get the current server load in cPanel in the "Serice Status" bit there is a Server Load section.

[IronWarrior]

Ah wish I know how to do that GatorDaveC.

Ah calum, the shared server am on, doesn't have "Service Status" wish it did though.

[calum]

Does your cPanel use the x2 or x3 skin?

Both of mine had it anyway.

Its down the left hand side next to all the details about bandwidth and space usage, its near the bottom.

[IronWarrior]

I know where it is, I been though this before (lol), just not enabled for us. :P

[GvilleRick]

Seems I remember that you have SSH access. You can log in using putty or another SSH program and run top to see the server loads or the uptime command listed previously.

[GatorDaveC]

Originally Posted by IronWarrior

Ah wish I know how to do that GatorDaveC.

Ah calum, the shared server am on, doesn't have "Service Status" wish it did though.

PHP Code:

<?
echo `/usr/bin/uptime`;
?>

That'll do it sir.

[uberamd]

Just wondering, I noticed doing a full backup causes a spike in server load (as I view the graph currently it is at ~45 (wow, not caused by me), and when I did a full backup it spiked to 20), does this count against our CPU cycles as the system needs to compress everything which clearly is CPU intensive?

Is doing a full backup every night or every 2 nights going to cause problems?