On Fri, 17 Dec 2010 16:04:16 +0100 Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote: > On Fri, Dec 17, 2010 at 02:56:51PM +0000, Gerard Lally wrote: > > On Fri, 17 Dec 2010 15:19:30 +0100 > > Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote: > > > > > On Fri, Dec 17, 2010 at 01:57:48PM +0000, Gerard Lally wrote: > > > > Yes I did increase that substantially. > > > > > > how much ? > > > > I tried increasing it from the default 2048, first to 16384 and > > then to 32768. > > > > I also tried tuning TCP in /etc/sysctl.conf: > > > > net.inet.tcp.recvbuf_auto=1 > > net.inet.tcp.sendbuf_auto=1 > > net.inet.tcp.sendbuf_max=16777216 > > net.inet.tcp.recvbuf_max=16777216 > > > > But the same thing happened all the time - mbuf usage climbed and > > climbed until the system froze. > > if altq is causing latency to increase, it's expected that sendbuf > will be raised up to sendbuf_max. Then a single TCP connection will > eat 8192 mbuf for its send queue This is useful information. My head is spinning from all the advice on tuning - should I tune? should I not tune? Would you recommend increasing the sendbux_max setting even further then? -- Gerard Lally
Attachment:
signature.asc
Description: PGP signature