On Thu, 11 May 2006, Amit Aronovitch wrote: > On 5/9/06, guy keren <[EMAIL PROTECTED]> wrote: > > > > On Tue, 9 May 2006, Shlomi Fish wrote: > > > > > > 3) Larger swap might be a good idea indeed - how much > > > > free/unpartitioned disk we have? > > > > > > Well we have the entire RAID array which is completely unused at the > > > moment. > > > (even for files). There are also some old partitions on /dev/sda and > > > /dev/sdb > > > which are unused, can be backed up to the IDE disk, and re-utilised as > > > swap. > > > > don't add swap - it'll just make the performance problem more severe.. > > > > Why? > Do you mean that with more swap the OS would load more processes, > reducing performance?
adding swap cannot increase performance - it's as simple as that. so don't add swap. at all. because we have a performance problem, not an "out of virtual memory" problem. by adding swap, you'll increase I/O and make it even more probable that the OOM killer will start running (and OOM here is out of physical RAM, not out of swap space). > What do you think the optimal swap size for eskimo should be (with > current 320M ram and for future, say 580M)? the optimal swap size is approximately 0 ;) now please leave the swap space alone. why do you feel the need to increase a resource that you already have too much of? we never got to a state where swap space was full or even close to being full. perhaps you need to read some kernel sources in order to be convinced that you're wrong with your perception of swap space and RAM. -- guy "For world domination - press 1, or dial 0, and please hold, for the creator." -- nob o. dy
