reiser4: fd -1 problem ?

reiser4: fd -1 problem ?

[Artur Makówka]

Hello, i've been investigating this problem for few weeks, and it seems 
reiser4 can be reason for this.

My apache process is from no clear reason suddenly peaking to 95% CPU 
usage and it stays that way until i kill it and start again.

It happens 6-7 times a day, i have hosting services on my server. I 
thought it was apache fault, but it strings shows this:

[pid 31610] munmap(0xb6f29000, 729088)  = 0
[pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE, 
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000
[pid 31610] munmap(0xb6f29000, 729088)  = 0
[pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE, 
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000


in endless loop, when this strange 'lock' happens.

is it a problem of fd -1 ? of course that could be also apache bug, that 
it can't handle wrong fd number, but why there is wrong fd number?

i ran reiser4.fsck and it didnt help.

i use kernel 2.6.12.6 with latest 2.6.12 reiser4 patch, as it seemed the 
most stable from recent releases.

What could be the problem here, and how can i repair it ? should i keep 
running fsck.reiser4 until it finds it ?

it doesnt happen very often, but like i said 6-7 times a day, but it 
varies, sometimes it more often sometimes almost 0.

thanks in advance for response. (this is probably not Apache issue, i 
already asked for this many many times)

Re: reiser4: fd -1 problem ?

[Vladimir V. Saveliev]

Hello

Artur Makówka wrote:
> Hello, i've been investigating this problem for few weeks, and it seems
> reiser4 can be reason for this.
> 
> My apache process is from no clear reason suddenly peaking to 95% CPU
> usage and it stays that way until i kill it and start again.
> 
> It happens 6-7 times a day, i have hosting services on my server. I
> thought it was apache fault, but it strings shows this:
> 
> [pid 31610] munmap(0xb6f29000, 729088)  = 0
> [pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE,
> MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000
> [pid 31610] munmap(0xb6f29000, 729088)  = 0
> [pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE,
> MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000
> 
> 
> in endless loop, when this strange 'lock' happens.
> 
> is it a problem of fd -1 ? 

-1 is fine because MAP_ANONYMOUS is set.

> of course that could be also apache bug, that
> it can't handle wrong fd number, but why there is wrong fd number?
> 

they are using MAP_ANONYMOUS, fd is ignored in that case.

> i ran reiser4.fsck and it didnt help.
> 

what did reiser4.fsck report? No corruptions found?

> i use kernel 2.6.12.6 with latest 2.6.12 reiser4 patch, as it seemed the
> most stable from recent releases.
> 
> What could be the problem here, and how can i repair it ? should i keep
> running fsck.reiser4 until it finds it ?
> 
> it doesnt happen very often, but like i said 6-7 times a day, but it
> varies, sometimes it more often sometimes almost 0.
> 
> thanks in advance for response. (this is probably not Apache issue, i
> already asked for this many many times)
> 

so far it does not look as reiser4 problem either.

Flex, can we have apache server running on reiser4 for test purposes?

Re: reiser4: fd -1 problem ?

[Artur Makówka]

Vladimir V. Saveliev wrote:
> Hello
> 
> Artur Makówka wrote:
>> Hello, i've been investigating this problem for few weeks, and it seems
>> reiser4 can be reason for this.
>>
>> My apache process is from no clear reason suddenly peaking to 95% CPU
>> usage and it stays that way until i kill it and start again.
>>
>> It happens 6-7 times a day, i have hosting services on my server. I
>> thought it was apache fault, but it strings shows this:
>>
>> [pid 31610] munmap(0xb6f29000, 729088)  = 0
>> [pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE,
>> MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000
>> [pid 31610] munmap(0xb6f29000, 729088)  = 0
>> [pid 31610] mmap2(NULL, 729088, PROT_READ|PROT_WRITE,
>> MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6f29000
>>
>>
>> in endless loop, when this strange 'lock' happens.
>>
>> is it a problem of fd -1 ? 
> 
> -1 is fine because MAP_ANONYMOUS is set.
> 
>> of course that could be also apache bug, that
>> it can't handle wrong fd number, but why there is wrong fd number?
>>
> 
> they are using MAP_ANONYMOUS, fd is ignored in that case.
> 
>> i ran reiser4.fsck and it didnt help.
>>
> 
> what did reiser4.fsck report? No corruptions found?
> 
>> i use kernel 2.6.12.6 with latest 2.6.12 reiser4 patch, as it seemed the
>> most stable from recent releases.
>>
>> What could be the problem here, and how can i repair it ? should i keep
>> running fsck.reiser4 until it finds it ?
>>
>> it doesnt happen very often, but like i said 6-7 times a day, but it
>> varies, sometimes it more often sometimes almost 0.
>>
>> thanks in advance for response. (this is probably not Apache issue, i
>> already asked for this many many times)
>>
> 
> so far it does not look as reiser4 problem either.
> 
> Flex, can we have apache server running on reiser4 for test purposes?
> 
> 

this happens only sometimes, and changing to latest 2.6.14 kernel and 
reiser4 patch didnt help.

i suppose it would be hard to simulate this, i get only few such errors 
per day and its pretty busy server (and i never get these at night which 
means the fact that its busy is important i guess)

no idea if its reiser4, but that was my thought... maybe not, i don't know

any suggestions to what i should do next , when changing to never 
kernel/fs didn't help? i will run fsck.reiser4 --build-fs tonight, but 
if this wont help i have no idea