Box 47 Running in Red

[amar]

For the past couple of weeks Box 47 has had server load running in the red pretty much all day. As a result my sites are very slow. Please fix.

[grant]

Call support, or start a ticket, and they should remedy the problem pretty quickly. I had a problem last week with the server load status being around 60%. I called support and they told me that it was another customer on the same box as me running programs they ought not to be running. Problem was solved in about 2 minutes, and everything returned to normal.

[Early Out]

Quick, dumb question - when I look at the box stats, the Server Load is shown as, for example, "4.67 (4 cpus)." What are those numbers telling me?

[grant]

Quick, dumb question - when I look at the box stats, the Server Load is shown as, for example, "4.67 (4 cpus)." What are those numbers telling me?

I would like to know that as well. All I do know is that once it gets around 12% it's in the red. My numbers usually run about where yours are.

[alligosh]

the Server Load is shown as, for example, "4.67 (4 cpus)." What are those numbers telling me?

Those numbers are the Load Average for the server.

having said that, it is hard to explain. it isn't a percentage of CPU or any concrete number. It is derived from several factors, but most heavily on the number of processes currently running or waiting on the CPU. Of course, like many explainations in here, that is an over-simplification.

Basically, the lower the number, the less the server is doing at the given moment. Also remember that it is an average over a minute, five minutes, or fifteen minutes, depending on which number is reported by cpanel (I think it is the one minute stat, but don't quote me on that; I use the CLI and not cpanel for things like that). We like to keep that number between 5 and 10. A 1 is the ideal in a perfect world, but rarely seen in reality or production.

Also rememeber that a 12 or 15, or even 25, is not a bad thing as long as it isn't consistent or happens for too long. That's the real problem with a 1 minute Load Average. It isn't long enough to determine long term issues, nor is it smal enough to tell immediate on-the-processor statistics. It is a helpful number to say that I should look into stuff, or not, but is only one indicator.

Add to that the fact that a unix Load Average only tells you something if you know how the box is supposed to run, and how it does under normal situations. Otherwise, it is very misleading.

I guess my soapbox is just to say, use it as a watermark if you want, but back up what you think it means with other data and tests. Oh, and please don't let any of the cpanel red lights sway your opinions. They are not always accurate (aka, show a red when something is actually working, etc)

[Early Out]

Those numbers are the Load Average for the server.

having said that, it is hard to explain.....Let me see if I can summarize: the server stats numbers don't really mean anything, and even the red dots don't tell you anything worth knowing. :D

Actually, I understand fairly well. I used to manage a small LAN, just 200 users. But this was a decade ago, so we're talking about a small room full of servers, with piles of cables behind them that only a few of us could actually trace ("Yeah, you can cut that one - that's left over from when we finally ditched the token ring!"). Working with the machines every day, you get to know what's normal, and what's a sign of trouble. You can't define it for someone else - you just get a feel for it. I could tell, for example, just by glancing at one of our hub stacks, or at the lights on a RAID tower, if things were behaving normally.