load avg meaning.

[Down2TheC]

The load avg that's shown all over... what's the meaning?

Let's say the output from a "uptime" command.

07:41:32 up 34 days, 10:48, 2 users, load average: 1.73, 1.70, 1.63

I assume it's a bit goofed due to the VPS. So maybe that's 173% of my VM's resources?

Historically it runs at 0.03-0.15 and things start crawling when it's over 1.5
These are the same numbers seen in Virtuozzo and Plesk.

[Ron]

The load average is simply the number of runnable processes in the queue. On our system it also includes the number of processes waiting for I/O.

I don't know what a good number is for a VPS, nor how the VPS software accounts for shared I/O resources.

[Vin DSL]

Rule of thumb...

1'ish to 4'ish is normal.

Over 10, it starts becoming problematic.

Lag will become noticeable, 20's and above.

Servers become unresponsive in the 60-100 range.

Over 100 usually indicates a crash is imminent!

Personally, when I see numbers above the teens, I launch a ticket, and ask them to check it.

[thelucky4]

Rule of thumb...

1'ish to 4'ish is normal.

Over 10, it starts becoming problematic.

Lag will become noticeable, 20's and above.

Servers become unresponsive in the 60-100 range.

Over 100 usually indicates a crash is imminent!

Personally, when I see numbers above the teens, I launch a ticket, and ask them to check it.

Umm those #s are generally for dedicated, shared/reseller type of environments no? I believe most VPS (single allocated cpus) only go up to about 1.5 max. I've done builds of apache etc and that pushes it to about 1.2~1.3 and majority of system is unresponsive.

[Ron]

Yes, as usual Vinnie is being ego centric.

[Ron]

Yes, as usual Vinnie is being ego centric.

Then again with three of him, there are several points of view through which to wade.

[Down2TheC]

Umm those #s are generally for dedicated, shared/reseller type of environments no? I believe most VPS (single allocated cpus) only go up to about 1.5 max. I've done builds of apache etc and that pushes it to about 1.2~1.3 and majority of system is unresponsive.

That's what I'm seeing and it's not my procs doing it so it has to be for the node. This morning I had a top 10 showing 3.17 load avg and 100% idle for my procs, and the appropriate 0%CPU for all my procs listed. But yeah... anything over 1.5 and even logging into SSH is a slow process.

So I don't ever wanna see what 10 is like. Yikes!

[Gwaihir]

It depends a lot on the system specs.

If you have a system with four CPU cores, a queue depth of 4 means just one proces per CPU, which is no problem at all. That same four one a single core system is quite noticeable.

That's assuming the wait if for CPU power, which isn't always the case. The wait for I/O complicates the picture further. (Top can tell these two apart.)

So, don't go by the number. Basically all it tells you is that IF the machine feels sluggish AND the number is considerably higher than you're used to, you've got something to investigate further (starting with top, usually).

[ethix]

That's what I'm seeing and it's not my procs doing it so it has to be for the node.

if you let support know they are normally pretty quick in fixing the issue

[Down2TheC]

True. Since they put my node on a monitor, it's been happening less. Not sure if they're out there helping the account that's going nuts, but the other day they answered up within minutes and found someone's mysqld running away. So props to the support setup here.

[mitx]

Rule of thumb...

1'ish to 4'ish is normal.

Over 10, it starts becoming problematic.

Lag will become noticeable, 20's and above.

Servers become unresponsive in the 60-100 range.

Over 100 usually indicates a crash is imminent!

Personally, when I see numbers above the teens, I launch a ticket, and ask them to check it.

Your right because my loads are at 170+ sometimes. I didnt crash but it runs very slow. Its orbiter dedicated so I had to upgrade. I can't wait for my new machine to be setup now.