N40L/OI/Napp-it High Processor Loads when idle

[Wolfsbane2k]

Hi,
I've got a N40L microserver running OI 151a17 from a 8GB flash drive with 16GB ECC ram and 4x 3TB WD30EFRX NAS Drives, and having no end of problems with processor loading on the box going by sluggish response and the gnome-system-monitor logs showing between a 50% and a 100% load on average.

It's running Serviio (java) , it's sharing mirror1 and i'm connected to it with VNC to try and find the problem: "top" is showing the following after a bit of media streaming,

last pid: 16564; load avg: 4.64, 4.22, 3.51; up 0+02:36:54 01:42:48
128 processes: 127 sleeping, 1 on cpu
CPU states: 41.9% idle, 42.9% user, 15.3% kernel, 0.0% iowait, 0.0% swap
Kernel: 3227 ctxsw, 9310 trap, 1220 intr, 30184 syscall, 6 fork, 7340 flt
Memory: 16G phys mem, 8626M free mem, 512M total swap, 512M free swap

PID USERNAME NLWP PRI NICE SIZE RES STATE TIME CPU COMMAND
16560 gdm 1 55 0 15M 10M sleep 0:00 2.80% metacity
16562 gdm 1 55 0 14M 8780K sleep 0:00 2.25% metacity
12 root 28 59 0 12M 11M sleep 13:57 2.09% svc.configd
5983 gdm 1 59 0 7560K 6484K sleep 0:24 1.07% gconfd-2
10 root 14 59 0 8132K 6348K sleep 4:15 0.92% svc.startd
5861 noaccess 1 59 0 48M 43M sleep 0:35 0.89% Xvnc
5986 gdm 1 59 0 7624K 6496K sleep 0:25 0.48% gconfd-2
5919 gdm 2 53 0 15M 10M sleep 0:07 0.41% gnome-session
845 root 1 29 10 13M 12M sleep 1:14 0.35% perl
16555 root 1 59 0 3960K 2208K cpu/0 0:00 0.34% top
5927 gdm 2 53 0 15M 10M sleep 0:07 0.33% gnome-session
6024 gdm 1 59 0 8760K 7048K sleep 0:08 0.23% at-spi-registry
13273 serviio 47 47 4 237M 233M sleep 12:13 0.23% java
914 Tim 3 59 0 197M 77M sleep 0:38 0.21% Xorg
5913 gdm 1 59 0 3156K 2024K sleep 0:04 0.19% dbus-daemon
5926 gdm 1 59 0 3156K 2020K sleep 0:04 0.15% dbus-daemon
458 root 29 59 0 7388K 4464K sleep 0:30 0.10% nscd
6051 gdm 1 59 0 82M 18M sleep 0:02 0.09% gnome-power-man
6045 gdm 1 59 0 82M 18M sleep 0:02 0.08% gnome-power-man
884 zfssnap 2 59 0 14M 5372K sleep 0:16 0.08% time-sliderd
6052 gdm 1 59 0 88M 25M sleep 0:03 0.07% gdm-simple-gree
1505 gdm 1 59 0 96M 33M sleep 0:14 0.07% gdm-simple-gree
6011 gdm 1 59 0 75M 13M sleep 0:02 0.06% metacity
6048 gdm 1 59 0 88M 25M sleep 0:01 0.04% gdm-simple-gree
13983 root 205 29 10 53M 49M sleep 0:00 0.04% zdb
6009 gdm 1 59 0 94M 30M sleep 0:01 0.04% gnome-settings-
6023 gdm 1 59 0 94M 30M sleep 0:01 0.03% gnome-settings-
6008 gdm 1 59 0 7800K 5984K sleep 0:00 0.03% at-spi-registry

ZFS recordsize is 128K.

Is this something i've just got to put up with, given the N40's processor?

Ta

 

[_Gea]

you may check whats possible when
- using a server/text version of OI
- stop napp-it (/etc/init.d/napp-it stop) -> start/restart

Is your 8GB flash a SSD or USB stick. The second is bad.

 

[jwcalla]

The problem here is that your top is showing mostly desktop-oriented processes that are probably related to the VNC session, so it's hard to tell what the big culprit is.

The second thing is that you appear to have several sessions going on here... Or maybe OI top displays threads as processes so it's not an issue, but e.g. there are three metacity window managers listed and the GDM greeter comes up three times as well. A ps -leaf | grep metacity can help identify if these are related to a single common parent process or are in fact their own instances.

Can you log out of all user desktop sessions and just have it in an ordinary idle state and then log in via ssh and do a top?

However if you want to have a desktop session available for login, you might have to accept that it would entail some moderate CPU usage.

 

[jwcalla]

Another thing... I wouldn't rely on gnome-system-monitor to check system load, as unfortunately this program can really pound the hell out of a low-end processor... especially if it's updating the visual charts.

Instead, use uptime or top to check the load averages.

 

[Wolfsbane2k]

Thanks for the input all. It looks like for some reason the access to Gnome from VNC was stable, but the local gnome terminal was fully screwed, dumping itself approx every 10 minutes.

Took 4 reboots to clear it, but disappeared now.

I had just changed the screen res for the VNC display, might have got its knickers in a twist.

 

[Wolfsbane2k]

you may check whats possible when
- using a server/text version of OI
- stop napp-it (/etc/init.d/napp-it stop) -> start/restart

Is your 8GB flash a SSD or USB stick. The second is bad.

USB stick. Yes, it's slow, but felt it was reasonable for a low write machine.

(and noatime, nodiratime is set on it)

 

[_Gea]

USB stick. Yes, it's slow, but felt it was reasonable for a low write machine.

(and noatime, nodiratime is set on it)

Some operating systems like FreeNAS or SmartOS (Illumos based like OI) are optimized for such usage. OI is not - especially the live edition. I'am quite sure your performance or CPU usage may be much lower with the cheapest Sata SSD > 30 GB

(8 GB is too low because its nearly full, very bad situation for ZFS and any OS)

 

[Wolfsbane2k]

Some operating systems like FreeNAS or SmartOS (Illumos based like OI) are optimized for such usage. OI is not - especially the live edition. I'am quite sure your performance or CPU usage may be much lower with the cheapest Sata SSD > 30 GB

(8 GB is too low because its nearly full, very bad situation for ZFS and any OS)

Yep, 8GB is too small, but i broke my 16GB so didn't have anything floating around spare bigger, and i wanted to try it all out.

For me it's a question of loosing 1 of my spare 2 remaining SATA connections for SSD ( which i am hoping to use for 2x 1TB Drives shortly) or using a specifically designed internal USB port.

It's working currently, but if it starts breaking again i'll probably jump on the SSD route and look at additional SATA add-in cards in the future.