Re: Second opinion: oom-killer

2008-01-17 Thread Vitaly
On Jan 17, 2008 3:53 PM, Shachar Shemesh <[EMAIL PROTECTED]> wrote: > Oren Held wrote: > > > > B. I can't see why disabling the swap would help to AVOID oomkiller? Swap > > should ENLARGE the available memory space; disabling swap might cause > > triggering oomkiller more frequently. Maybe I misund

Re: Second opinion: oom-killer

2008-01-17 Thread Shachar Shemesh
Oren Held wrote: B. I can't see why disabling the swap would help to AVOID oomkiller? Swap should ENLARGE the available memory space; disabling swap might cause triggering oomkiller more frequently. Maybe I misunderstood what you meant. The memory I suspect the system is running out of is t

Re: Second opinion: oom-killer

2008-01-17 Thread Oren Held
I have no direct answer to your question. Oom-killer should be triggered in a memory emergency, and not when there's enough free memory space. Some notes though: A. oom killer doesn't randomly choose a process to kill.. I described the algorithm for choosing a candidate for killing in: http:/

Re: Second opinion: oom-killer

2008-01-17 Thread Gilad Ben-Yossef
Shachar Shemesh wrote: Gilad Ben-Yossef wrote: - As a first stage, disable the swap. This doesn't make much sense to me. What is it suppose to achieve? It is SUPPOSED to achieve less memory to keep track of. As far as I understand it, the kernel keeps track over which virtual pages reside i

Re: Second opinion: oom-killer

2008-01-17 Thread Shachar Shemesh
Gilad Ben-Yossef wrote: - As a first stage, disable the swap. This doesn't make much sense to me. What is it suppose to achieve? It is SUPPOSED to achieve less memory to keep track of. As far as I understand it, the kernel keeps track over which virtual pages reside in which physical locati

Re: Second opinion: oom-killer

2008-01-17 Thread Gilad Ben-Yossef
Shachar Shemesh wrote: Machine with 16GB ram and a bit of swap, running 32bit gentoo with (as far as I know) 3:1 memory split. The symptoms: Every so often the oom-killer kicks in, for no apparent reason. The processes it kills appear to be randomly chosen. Monitoring was not turned on,

Re: Second opinion: oom-killer

2008-01-17 Thread Gilboa Davara
On Thu, 2008-01-17 at 12:32 +0200, Shachar Shemesh wrote: > Hi list, > > > I need the esteemed people's second opinion to a recommendation I gave a > client. > > > The setup: > > Machine with 16GB ram and a bit of swap, running 32bit gentoo with (as > far as I know) 3:1 memory split. > >

Second opinion: oom-killer

2008-01-17 Thread Shachar Shemesh
Hi list, I need the esteemed people's second opinion to a recommendation I gave a client. The setup: Machine with 16GB ram and a bit of swap, running 32bit gentoo with (as far as I know) 3:1 memory split. The symptoms: Every so often the oom-killer kicks in, for no apparent reason. The