On Monday 30 April 2007 01:54:46 pm Lars Engels wrote: > On Mon, Apr 30, 2007 at 01:36:41PM -0400, John Baldwin wrote: > > On Monday 30 April 2007 01:02:04 pm Lars Engels wrote: > > > On Mon, Apr 30, 2007 at 12:08:58PM -0400, John Baldwin wrote: > > > > On Friday 27 April 2007 04:57:52 pm Lars Engels wrote: > > > > > On Thu, Apr 26, 2007 at 05:45:50PM -0400, John Baldwin wrote: > > > > > > > > > > I also encountered occasional "Panic: failed to create swap zone" when > > > > > CPU #1 was launched just before the disks get mounted. Could that be > > > > > related? > > > > > > > > Ok, so for the default setup, the only problem is that panic? That sounds > > > > like running out of kvm possibly as the swapzone is preallocated from > > kmem. > > > > The system may be allocating too much for the swapzone, in which case you > > can > > > > explicitly set the size via the 'kern.maxswzone' tunable. One possible > > > > formula for this is: > > > > > > > > maxswzone = (swap + 1024) * 1024 * 9 / 2 > > > > > > > > Where 'swap' is the amount of swap in megabytes. > > > > > > I added kern.maxswzone=13953024 (2004 MB swap) to loader.conf but after > > > a reboot the oid is still unknown. Do I need to set it somewhere else? > > > > Unfortunately, it's just a tunable, there isn't a sysctl to see the current > > setting. You can, however, look at the 'vmstat -z' output. You can figure > > out the current swap zone size by multiplying the size and limit columns > > for 'SWAPMETA'. Something like this: > > > > vmstat -z | awk '/^SWAPMETA/ { printf "%d\n", $2 * $3 }' | bc > > There's a difference of 120 bytes between the vmstat output and > maxswzone. Is that normal? Yes. The swapzone objects are about ~280 bytes or so. > But I think the panics are gone now. I booted five times with the > cardbus card inserted when cpu #1 was launched and there was no panic. Ok. -- John BaldwinReceived on Mon Apr 30 2007 - 16:34:14 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:09 UTC