> I switched rufus.w3.org to 2.2.1, the good point is that I didn't
> experience any stability problem compared to 2.2.0 . However I noticed
> a substancial shortage of memory, the exact same setup running under
> 2.0.35 used to have around 40Mbytes (of the 128MB of main memory) used
> for buffer cache and nearly no swap used at all.
>
> After 24 hours under the same load and with the exact same OS/software
> setup but under 2.2.1, I have only around 11 MB allocated to the cache
> and 40 MBytes of swap used. The machine is swapping a lot. There is around
> 70 ftpd and 30 httpd running in average. Is there anything which could explain
> an increase of 300 KBytes of virtual memory usage per daemon ?
Actually, a couple of hours later the swap had grown to 70MBytes without
any real differences in the number of request served. The machine was slowly
going to an halt so I saved a copy of most /proc files for post analysis
and rebooted.
My feeling is that I have hit some kind of memory leak. The only strange
points of my setup are that I'm using IDE drives hooked to the on-board
IDE interface: Acer Labs M5229 TXpro, and a Promise Technology IDE UltraDMA/33
This is not a SCSI based system. Ethernet is an EEPro 100.
The only strange points logged by the kernel were around 15 messages like:
hdf: dma_intr: status=0x51 { DriveReady SeekComplete Error }
hdf: dma_intr: error=0x84 { DriveStatusError BadCRC }
I can provide more informations if needed,
Daniel
--
[Yes, I have moved back to France !]
[EMAIL PROTECTED] | W3C INRIA Rhone-Alpes | Today's Bookmarks :
Tel : +33 476 615 257 | 655, avenue de l'Europe | Linux, WWW, rpmfind,
Fax : +33 476 615 207 | 38330 Montbonnot FRANCE | rpm2html, Kaffe,
http://www.w3.org/People/W3Cpeople.html#Veillard | badminton, and Amaya.
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
