Hi all,
somewhere in the 2.4.21-xx series (currently running -rc1 here) seems
to be a VM regression, that results in quite hefty swap-outs. The box
is an Athlon-XP with 512MB RAM, running batched gcc jobs (distribution
builds taking three days and more...). But the box is used a normal
workstation, too. My normal desktop programs constantly need to get
swapped-in - quite annoying. This only start to happen after an uptime
of a day.
So the CPU:
vendor_id : AuthenticAMD
cpu family : 6
model : 6
model name : AMD Athlon(tm) 4 Processor
stepping : 2
cpu MHz : 1460.481
cache size : 256 KB
$ cat /proc/meminfo
total: used: free: shared: buffers: cached:
Mem: 527351808 519213056 8138752 0 88793088 328437760
Swap: 537214976 184963072 352251904
MemTotal: 514992 kB
MemFree: 7948 kB
MemShared: 0 kB
Buffers: 86712 kB
Cached: 252328 kB
SwapCached: 68412 kB
Active: 276828 kB
Inactive: 177468 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 514992 kB
LowFree: 7948 kB
SwapTotal: 524624 kB
SwapFree: 343996 kB
$ uptime
21:52:41 up 4 days, 23:55, 4 users, load average: 2.14, 2.47, 2.32
$ cat /proc/version
Linux version 2.4.21-rc1 (root@jackson) (gcc version 3.2.2) #1 Mon Apr 28 00:36:17 CEST 2003
More data available on request.
- Ren?
--
Ren? Rebe - Europe/Germany/Berlin
e-mail: [email protected], [email protected]
web: http://www.rocklinux.org/people/rene http://gsmp.tfh-berlin.de/rene/
Anyone sending unwanted advertising e-mail to this address will be
charged $25 for network traffic and computing time. By extracting my
address from this message or its header, you agree to these terms.
Hi,
2.5.68 is much smoother, quite less swap is used, my editors are not
pulled out of the swap all the time:
$ free
total used free shared buffers cached
Mem: 515848 505804 10044 0 92808 173876
-/+ buffers/cache: 239120 276728
Swap: 524624 35612 489012
Linux version 2.5.68 (root@jackson) (gcc version 3.2.2) #2 Sun May 4 00:20:13 CEST 2003
20:39:52 up 1 day, 20:17, 2 users, load average: 3.31, 3.36, 3.19
On: Sat, 03 May 2003 21:55:34 +0200 (CEST),
Rene Rebe <[email protected]> wrote:
> Hi all,
>
> somewhere in the 2.4.21-xx series (currently running -rc1 here) seems
> to be a VM regression, that results in quite hefty swap-outs. The box
> is an Athlon-XP with 512MB RAM, running batched gcc jobs (distribution
> builds taking three days and more...). But the box is used a normal
> workstation, too. My normal desktop programs constantly need to get
> swapped-in - quite annoying. This only start to happen after an uptime
> of a day.
>
> So the CPU:
>
> vendor_id : AuthenticAMD
> cpu family : 6
> model : 6
> model name : AMD Athlon(tm) 4 Processor
> stepping : 2
> cpu MHz : 1460.481
> cache size : 256 KB
>
> $ cat /proc/meminfo
> total: used: free: shared: buffers: cached:
> Mem: 527351808 519213056 8138752 0 88793088 328437760
> Swap: 537214976 184963072 352251904
> MemTotal: 514992 kB
> MemFree: 7948 kB
> MemShared: 0 kB
> Buffers: 86712 kB
> Cached: 252328 kB
> SwapCached: 68412 kB
> Active: 276828 kB
> Inactive: 177468 kB
> HighTotal: 0 kB
> HighFree: 0 kB
> LowTotal: 514992 kB
> LowFree: 7948 kB
> SwapTotal: 524624 kB
> SwapFree: 343996 kB
>
> $ uptime
> 21:52:41 up 4 days, 23:55, 4 users, load average: 2.14, 2.47, 2.32
>
> $ cat /proc/version
> Linux version 2.4.21-rc1 (root@jackson) (gcc version 3.2.2) #1 Mon Apr 28 00:36:17 CEST 2003
>
> More data available on request.
>
> - Ren?
- Ren?
--
Ren? Rebe - Europe/Germany/Berlin
e-mail: [email protected], [email protected]
web: http://www.rocklinux.org/people/rene http://gsmp.tfh-berlin.de/rene/
Anyone sending unwanted advertising e-mail to this address will be
charged $25 for network traffic and computing time. By extracting my
address from this message or its header, you agree to these terms.
On Mon, May 05, 2003 at 08:40:36PM +0200, Rene Rebe wrote:
> Hi,
>
> 2.5.68 is much smoother, quite less swap is used, my editors are not
> pulled out of the swap all the time:
>
> $ free
> total used free shared buffers cached
> Mem: 515848 505804 10044 0 92808 173876
> -/+ buffers/cache: 239120 276728
> Swap: 524624 35612 489012
You will notice that -aa is better and -rmap is even better still.
I can run badblocks on two drives without very much swapping on 192MB with
-rmap15[aeg]