2.2.17-RAID: Eating memory

2.2.17-RAID: Eating memory

[John Plate]

Hi

I have patched the 2.2.17 kernel (Debian) with the "new RAID" patches
with no errors.

I experienced a the following problem:

Two partitions are set up as RAID-1, but not yet fully
created/initialized. On boot, the process starts/continues
automatically (partitions /dev/hda3 and /dev/hdc3 set to type "Linux
raid autodetect", but not mounted in /etc/fstab).

/proc/mdstat says: 
md2 : active raid1 hdc3[1] hda3[0] 4723008 blocks [2/2] [UU]
resync=11% finish=77.3min

While this is active, I start an rsync between two other partitions
(/dev/hda1 and /dev/hdc1). This should be fine because RAID is
expected to work in background only. After a while the system reports
on the console a lot of "VM: do_try_to_free_pages failed" for many of
the daemon programs running. Before that, the swap usage increases but
there is enough swap space.

I guess that this is some kind of race condition because the RAID
program runs fine with no other heavy load. The rsync also behaves
well under other circumstances. I expect the problem to be with the
RAID code. 

The system is stable, it has never crashed. This is my first problem
with a Linux kernel.

Can anyone please help?
-- 
John Plate 

Re: 2.2.17-RAID: Eating memory

[Francois Romieu]

John Plate <plate@infotek.dk> écrit :
[do_try_to_free_pages]

Known issue on < 2.2.19pre kernels (do a search for "do_try_to_free_pages"
on the ML archive). It doesn't require RAID btw. Upgrade to 2.2.19prexx, it
performs better.

-- 
Ueimor