Without fastcgi

[Z4P]

Hmm jeay me too.

@Joximu, Thanks all small bits will help.

[ephigenie]

I don't think that in the end ispCP needs more Ram than vhcs2.

Sometimes i might even be a bit less - because of the use of mpm-worker which consumes less memory altogether because it uses a threaded worker which doesn't have to fork a process for a requestslot like the prefork worker does.

so there're only the php-processes left... They actually can consume a lot of ram - but i don't think you're running 200+ domains on a 128Mb vserver so its around 15-30Megs of Ram per Customer max used.

[joximu]

@z4p

well, you also can disable bind9 - as cube does.
Your DNS is/are managed somewhere else...

[ArcAiN6]

ok... is it just me, or does he not have any swp space at all?? THis could be a reason for poor server performance.
his reads:

Code:

Mem:    471856k total,    44396k used,   427460k free,        0k buffers
Swap:        0k total,        0k used,        0k free,        0k cached

mine reads:

Code:

Mem:    257108k total,   253236k used,     3872k free,    52292k buffers
Swap:   522104k total,   109028k used,   413076k free,    28864k cached

and mine seems to run pretty fast...

[BeNe]

SWAP is bad!!!

Why should his server create a swap ?
He has enough memory free!

[ephigenie]

Swap is bad for performance - but every server needs swapspace - even the 32Gb Sun Server have huge swap areas.

Swapping is the only way if there's too few Ram.

So on Systems with < 256 Megs Ram you really need swap!
On Vservers it seems a bit more complicated as i tried to let him create some swap but this seems to be managed in the upper layer of its paravirtualization. So maybe the limits set on that server are just too aggressive.

If he actually could use those 400+ Megs of Ram it would be more than enough.
But it seems as if he cannot use that Ram as he wants to.
And the virtualization - manager kicks those processes out of Ram that go above 128Mb

[Z4P]

Hmm, stil is apache crashing down alot. but strange because top gives i have enougf memory:

Code:

top - 01:34:17 up 3 days,  9:27,  1 user,  load average: 0.00, 0.00, 0.00
Tasks:  35 total,   2 running,  33 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0% us,  0.0% sy,  0.0% ni, 100.0% id,  0.0% wa,  0.0% hi,  0.0% si
Mem:    471856k total,    29728k used,   442128k free,        0k buffers
Swap:        0k total,        0k used,        0k free,        0k cached

[Daff]

Hm, got the same Problem. I installed OpenVZ on a server lately and now I have a nice Playground. But Apache doesn't even start when I set privvmpages to less than 512 Megs. For comparison:

My current System (about 40 Webpages) running VHCS at a VPS Hoster:

VPS Memory Usage
Currently used: 197.664 MB
Warranted: 512 MB
Burstable: 1536 MB

Now the freshly installed ISPCP Omega RC2c on my own Hardware Node:

VPS Memory Usage
Currently used: 634.734 MB
Warranted: 128 MB
Burstable: 1024 MB

I had to set
PRIVVMPAGES="262144:262144"
Just to get Apache running at all.
I basically wanted to create a VPS with 128 Megs RAM and 256 Megs Burstable with ISPCP for a new costumer (I now its too less but for more he simply has to pay more . So its not very economic for me when he sees that he has 1 Gig Available (even if 896 Megs are swapped in fact).

[ephigenie]

Hm this is bad - way too much memory is used

I wonder which part takes it.
On my physical servers it doesn't look like that - perhaps there's some shm allocated somewhere. Perhaps you can try to use mod_fcgid instead of mod_fastcgi ?

How much memory do you allow your php to use ? 8,16,32,64 ?

[Daff]

Thats the basic installation config and the base system (no accounts, nothing changed in the config). So I think the default is 8 megs. I don't get the reason for such a huge memory consumption. The Apache error log gives this (with 512 Megs of RAM burstable!):

[Fri May 25 09:55:17 2007] [alert] (12)Cannot allocate memory: apr_thread_create: unable to create worker thread
[Fri May 25 09:55:17 2007] [alert] (12)Cannot allocate memory: apr_thread_create: unable to create worker thread
[Fri May 25 09:55:19 2007] [alert] No active workers found... Apache is exiting!

I disabled greylisting and if I stop bind apache starts but still produces a lot of this error messages. The base OpenVZ Configuration ist from my current VPS at http://www.vc-server.de (VDS Pro) so there should be enough ressources except the memory stuff. The PRIVVMPAGES seem to be the main problem which is the size of private (or potentially private) memory allocated by an application.