server load techie question (top)

[fusionx]

CPU states: 7.0% user, 0.0% nice, 91.9% system, 1.1% interrupt, 0.0% idle


In Top, does %idle mean the same thing as in Windows? Does this mean the CPU is maxed?

[SoundMan]

91.9% system is not good.

[VeriSexy]

last pid: 58913; load averages: 0.21, 0.46, 0.59 up 142+20:06:25 18:23:16
148 processes: 1 running, 147 sleeping
CPU states: 3.9% user, 0.0% nice, 4.3% system, 0.8% interrupt, 91.1% idle
Mem: 337M Active, 1147M Inact, 164M Wired, 48M Cache, 31M Buf, 313M Free
Swap: 2048M Total, 4K Used, 2048M Free

[infecto]

I hope to god you are running a few scripts (as in building a site with tgp script or something like that) while you checked that....if not thats really not good lol.

[Los]

do a ps -aux to see what is running or top and see whats killing your cpu

[fuzebox]

Weird to see system load that high, and not user... Are you swapping maybe?

[Sami]

Yes that means your CPU is maxed out.


If you have a site trading script thats using mysql and its not optimized, they usually cause that.

But like it was mentioned before,
do a ps -axf

to see whats going on.

[fusionx]

Unfortunately, I'm on a virtual server - shared host.

::Rant mode on::

I called this morning (9 am EST) and reported a slow site - cgi-bin stuff was crawling. Top showed 0.0% idle. There are several TAR processes running sucking up all the CPU. They said they were running backups.

I checked again at about 9:00 PM, and sure enough, still slow. The same TAR processes are still running. I called back and expressed my disbelief that a tar backup would cause this. They then said they were moving some sites off the server and this might be why.. might.. the tech did say he'd check into it, and seemed concerned.

It's 1:22 AM and the same TAR processes are still running, and the site response absolutely sucks.

I'm no unix admin, but at least I know about Nice. Sheesh.. talk about a wasted day. Wanted to set up two more sites and can't even modify a template in Autogallery SQL.

Anyway....

::rant mode off::

Thanks for the responses!

[Mr.Fiction]

Post the processes listed when you run top and your memory stats from top.

[fusionx]

Originally posted by Mr.Fiction
Post the processes listed when you run top and your memory stats from top.

last pid: 24662; load averages: 7.66, 9.48, 9.89 up 63+19:00:14 01:39:20
318 processes: 8 running, 303 sleeping, 7 zombie
CPU states: 4.0% user, 0.0% nice, 94.7% system, 1.3% interrupt, 0.0% idle
Mem: 1243M Active, 295M Inact, 396M Wired, 72M Cache, 199M Buf, 3252K Free
Swap: 4079M Total, 25M Used, 4054M Free

PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU CPU COMMAND
53735 root 64 0 528K 80K CPU3 3 244.1H 91.41% 91.41% tar
95513 root 64 0 528K 80K CPU0 0 232.4H 84.28% 84.28% tar
16509 root 63 0 528K 80K RUN 2 237.8H 45.56% 45.56% tar
4907 root 64 0 528K 80K CPU2 2 78.2H 43.65% 43.65% tar
96510 root 63 0 528K 80K RUN 1 234.7H 30.27% 30.27% tar
52792 root 62 0 528K 80K RUN 1 240.8H 28.96% 28.96% tar
28732 root 63 0 528K 80K RUN 1 229.7H 28.32% 28.32% tar
24658 fusionx 50 0 2280K 1492K CPU1 2 0:02 37.13% 6.74% top
24660 www 2 0 3648K 3172K kqread 0 0:00 4.20% 0.59% perl
24153 www 2 0 76728K 11936K sbwait 3 0:00 0.34% 0.34% httpd
24022 www 2 0 73392K 9320K accept 0 0:00 0.05% 0.05% httpd
24167 www 2 0 73060K 9572K poll 2 0:00 0.05% 0.05% httpd
24266 www 2 0 72604K 8480K kqread 1 0:00 0.05% 0.05% httpd
24341 www 2 0 72208K 8352K sbwait 2 0:00 0.05% 0.05% httpd
24659 www 2 0 2096K 1620K kqread 2 0:00 0.35% 0.05% perl
88786 root 10 0 2644K 640K nanslp 0 74:02 0.00% 0.00% daemon
68078 root 10 0 2644K 640K nanslp 0 52:18 0.00% 0.00% daemon

[webair]

fusionx

contact me I have a solution that we can upgrade you at no additional cost to alleviate this issue...

[Mr.Fiction]

Originally posted by fusionx


last pid: 24662; load averages: 7.66, 9.48, 9.89 up 63+19:00:14 01:39:20
318 processes: 8 running, 303 sleeping, 7 zombie
CPU states: 4.0% user, 0.0% nice, 94.7% system, 1.3% interrupt, 0.0% idle
Mem: 1243M Active, 295M Inact, 396M Wired, 72M Cache, 199M Buf, 3252K Free
Swap: 4079M Total, 25M Used, 4054M Free

PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU CPU COMMAND
53735 root 64 0 528K 80K CPU3 3 244.1H 91.41% 91.41% tar
95513 root 64 0 528K 80K CPU0 0 232.4H 84.28% 84.28% tar
16509 root 63 0 528K 80K RUN 2 237.8H 45.56% 45.56% tar
4907 root 64 0 528K 80K CPU2 2 78.2H 43.65% 43.65% tar
96510 root 63 0 528K 80K RUN 1 234.7H 30.27% 30.27% tar
52792 root 62 0 528K 80K RUN 1 240.8H 28.96% 28.96% tar
28732 root 63 0 528K 80K RUN 1 229.7H 28.32% 28.32% tar
24658 fusionx 50 0 2280K 1492K CPU1 2 0:02 37.13% 6.74% top
24660 www 2 0 3648K 3172K kqread 0 0:00 4.20% 0.59% perl
24153 www 2 0 76728K 11936K sbwait 3 0:00 0.34% 0.34% httpd
24022 www 2 0 73392K 9320K accept 0 0:00 0.05% 0.05% httpd
24167 www 2 0 73060K 9572K poll 2 0:00 0.05% 0.05% httpd
24266 www 2 0 72604K 8480K kqread 1 0:00 0.05% 0.05% httpd
24341 www 2 0 72208K 8352K sbwait 2 0:00 0.05% 0.05% httpd
24659 www 2 0 2096K 1620K kqread 2 0:00 0.35% 0.05% perl
88786 root 10 0 2644K 640K nanslp 0 74:02 0.00% 0.00% daemon
68078 root 10 0 2644K 640K nanslp 0 52:18 0.00% 0.00% daemon

How often does your host run those backups? Is that daily?

[salsbury]

nice wouldn't help in this case. nice doesn't help with disk intensive processes - only with cpu-bound processes. what *would* help however, is for them to only run one tar at a time instead of 7. 7 makes the disk have to work awfully hard seeking all over the place.

i'm betting that you're on an IDE-based server. that'd explain the extremely high system CPU.

it might be time to find a host that uses SCSI exclusively. even on a shared server, you should see far fewer slowdowns from stuff like that.

[fusionx]

Originally posted by Mr.Fiction




How often does your host run those backups? Is that daily?

Nah - I think it's a one time process. I've never had a day where response was slow all day long.. a moment here and there - no problem.

[fusionx]

CPU states: 0.9% user, 0.0% nice, 3.1% system, 0.4% interrupt, 95.6% idle


Ahhh... perfection

Webair Mike just got it hooked up. That's why I love Webair. Sometimes they have issues, like every hosting company, but they always get it fixed. Must be a new tech :-)