Topic: Server Crashes

[arnoldob]

Hi,
I currently have SME 7.1.3 running on:
MSI K8T NEO2-F Motherboard (Via K8T800 chipset)
AMD Athlon64 3000+
1Gb DDR 400 memory
Dual Seagate 160GB IDE drives in software raid 1

I keep having problems with SME crashing. I began running htop to monitor the server, when it last crashed htop was showing 100% CPU, 100% memory, 100% swap files usage, and of course the putty session i was using for this was unresponsive. There were 462 processes vs a normal usage pattern of 180 to 200. The majority of the additional processes were apache and mysql beating the life out of the server for a wordpress website.

Eventually I'll move the offending website to it's own server so that the e-mail and other stuff doesn't die with it. For now when situations like this occur how should it be handled? Once it gets this far I have tried unplugging the WAN cable. I would expect it to recover after a few minutes, but the admin console and local tty sessions remain unresponsive. Is that normal behavior? Thus far the only thing that seems to work is pressing the hardware reset button, but that seems rather... primative and dangerous to the file system. If it maxes out swap, cpu and memory should I expect SME to recover gracefully?

[dmay]

For now when situations like this occur how should it be handled?

Identify the roque process(es) and take them down. Wait a minute or two. Restart. Most of SME key services are run by daemontools. See /service and this:

http://cr.yp.to/daemontools/svc.html

Darrell

[CharlieBrady]

For now when situations like this occur how should it be handled?

Identify the roque process(es) and take them down. Wait a minute or two. Restart. Most of SME key services are run by daemontools. See /service and this:

http://cr.yp.to/daemontools/svc.html

SME7 uses runit rather than daemontools. Very similar, but not identical. See:

http://smarden.org/runit/
 
Use the sv command to control processes:

http://smarden.org/runit/sv.8.html

Since poster reports swap 100% used, then a memory leak somewhere is a possibility.

[Tib]

I have a similar problem with ntop .... on my work machines ntop runs fine but on my home machine ntop uses 100% of the mem.

All i do for now is /etc/init.d/ntop stop and everything goes back to normal.

Maybe you should just kill htop and see if all is ok.

Then try and figure out why htop is causing problems ... if htop is the problem.

I will in the future uninstall ntop and just use top.

Regards,

Tib

[arnoldob]

I don't have an issue with htop. It seems to use 1-2% of CPU when it's running. The server came to a standstill with or without htop running.