All of lore.kernel.org
 help / color / mirror / Atom feed
From: Markus Wenke <M.Wenke@web.de>
To: Eric Dumazet <dada1@cosmosbay.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>, linux-kernel@vger.kernel.org
Subject: Re: to many sockets ?
Date: Thu, 05 Oct 2006 11:16:17 +0200	[thread overview]
Message-ID: <4524CD61.5080904@web.de> (raw)
In-Reply-To: <200610051045.46919.dada1@cosmosbay.com>

Eric Dumazet schrieb:
> On Thursday 05 October 2006 10:20, Markus Wenke wrote:
>   
>> Eric Dumazet schrieb:
>>     
>
>   
>>> Could you post here the result of these commands when your system is
>>> using more than 100.000 connections (and before the OOM :) )
>>>       
>> Hi,
>>
>> here the results with 130001 connetions
>>
>>     
>>> cat /proc/meminfo
>>>       
>> MemTotal:      3108372 kB
>> MemFree:       2114404 kB
>> Buffers:          5112 kB
>> Cached:          97804 kB
>> SwapCached:          0 kB
>> Active:         140552 kB
>> Inactive:        38948 kB
>> HighTotal:     2228160 kB
>> HighFree:      2048108 kB
>> LowTotal:       880212 kB
>> LowFree:         66296 kB
>>     
>
> See here ? you have 'only' 880212 kB of LOWMEM, and 66 MB free.
> all kernel structures (you can see them in /proc/slabinfo) are lying on this 
> zone, no matter you add RAM on your machine (more RAM end up in HighMEM zone, 
> wich is basically unused on your setup)
>
> Since you have a 64bits CPU, your best move would be to use a 64bits kernel 
> (you can keep all user land in 32bits mode)
>
> With a 64bits kernel, kernel land structures would not be constrained in a 
> small area, but can use full RAM.
>
I think I change to the 64 Bit-Kernel ;-).


btw: with the Xen-patched Kernel I get this at 130000 connections:

MemTotal:      2979840 kB
MemFree:       2394544 kB
Buffers:          5748 kB
Cached:         104492 kB
SwapCached:          0 kB
Active:         146632 kB
Inactive:        43276 kB
HighTotal:     2136004 kB
HighFree:      1943728 kB
LowTotal:       843836 kB
LowFree:        450816 kB
SwapTotal:     2104472 kB
SwapFree:      2104472 kB
Dirty:             412 kB
Writeback:           0 kB
Mapped:          89556 kB
Slab:           285284 kB
CommitLimit:   3594392 kB
Committed_AS:   205244 kB
PageTables:        756 kB
VmallocTotal:   114680 kB
VmallocUsed:      4576 kB
VmallocChunk:   109280 kB

> I'm curious you have so many sockets but few entries in route cache... 
> basically all connections come from few machines ?
>
My test scenario has only 5 clients, so I must make the connections with 
a few clients.



Thanks a lot

Markus Wenke


  reply	other threads:[~2006-10-05  9:16 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-10-04 15:03 to many sockets ? Markus Wenke
2006-10-04 16:33 ` Alan Cox
2006-10-05  7:14   ` Markus Wenke
2006-10-05  7:58     ` Eric Dumazet
2006-10-05  8:20       ` Markus Wenke
2006-10-05  8:45         ` Eric Dumazet
2006-10-05  9:16           ` Markus Wenke [this message]
2006-10-05 21:30     ` Alan Cox
2006-10-06  5:32       ` Markus Wenke

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4524CD61.5080904@web.de \
    --to=m.wenke@web.de \
    --cc=alan@lxorguk.ukuu.org.uk \
    --cc=dada1@cosmosbay.com \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.