Skip to contentSkip to conversation

3.7.0-rc2 can't handle mutt (with 3.7G mail file) +FF (4 tabs) on a 4G memory+4 core system ?

headache Ortwin Glück 2 participants 7 messages Nov 8, 2012 – Nov 11, 2012

lore
Share & tools
lore
#0 lore raw

To me this looks like an issue with swap. Can you try without swap (swapoff)?

Ortwin

#1 lore raw
Quoted 2 lines · Ortwin Glück

To me this looks like an issue with swap. Can you try without swap (swapoff)?

Not sure but my random guess is it could be related to a small tmpfs in my setting

tmpfs 1869900 240 1869660 1% /tmp

As I just noticed that I couldn't quit from mutt due to tmpfs is full.

Maybe it was just shrinking tmpfs caused problem? then I can think of deleting tmpfs from my setup..

/l

#2 lore raw
Quoted 1 line · Luming Yu

As I just noticed that I couldn't quit from mutt due to tmpfs is full.

That's also pointing towards high memory pressure.

#3 lore raw
Quoted 8 lines · Ortwin Glück
Quoted 2 lines · Luming Yu

As I just noticed that I couldn't quit from mutt due to tmpfs is full.

That's also pointing towards high memory pressure.

watch "cat /proc/meminfo" just showed me the progress
mutt was creating a very large tmp file in /tmp/ until "memfree" reached a watermark then failed with a message like " no space in /tmp"

sounds like I can't use mutt if my mutt mail file is close to the physical memory
I have when tmp-on-tmpfs is enabled..

hmm....make sens? maybe we can't use tmp-on-tmpfs feature in f18

#4 lore raw
Quoted 14 lines · Luming Yu
Quoted 8 lines · Ortwin Glück
Quoted 2 lines · Luming Yu

As I just noticed that I couldn't quit from mutt due to tmpfs is full.

That's also pointing towards high memory pressure.

watch "cat /proc/meminfo" just showed me the progress
mutt was creating a very large tmp file in /tmp/ until "memfree" reached a watermark then failed with a message like " no space in /tmp"

Quoted 5 lines

sounds like I can't use mutt if my mutt mail file is close to the physical memory
I have when tmp-on-tmpfs is enabled..

hmm....make sens? maybe we can't use tmp-on-tmpfs feature in f18

but why not having disk to backup tmpfs under memory pressure?

#5 lore raw
Quoted 2 lines · Luming Yu
Quoted 1 line

hmm....make sens? maybe we can't use tmp-on-tmpfs feature in f18

but why not having disk to backup tmpfs under memory pressure?

It should. tmpfs should use all available RAM+swap. But then I guess you have no swap, and the kernel simply is desperately trying to find memory for a slab/slub or whatever. OOM killer can't help because the temp file uses the space, not a process.

#6 lore raw
Quoted 8 lines · Ortwin Glück
Quoted 4 lines · Luming Yu
Quoted 2 lines

hmm....make sens? maybe we can't use tmp-on-tmpfs feature in f18

but why not having disk to backup tmpfs under memory pressure?

It should. tmpfs should use all available RAM+swap. But then I guess you

there is swap space available, but somehow it didn't help as the attached pic shows in my latest testing with 3.7.0-rc4

Quoted 4 lines

have no swap, and the kernel simply is desperately trying to find memory for a slab/slub or whatever. OOM killer can't help because the temp file uses the space, not a process.