On Thu, May 23, 2013 at 01:21:55PM -0400, james wrote: > We had a short outage today and I think it might have been due to our > firewall. I can see that we hit the limit for this variable but I'm not > really sure what it does or if it is safe to set it higher. I found the > following at the time of the outage in the log: > > May 23 09:57:27 shiva2 /bsd: WARNING: mclpool limit reached; increase > kern.maxclusters > May 23 09:58:27 shiva2 /bsd: WARNING: mclpool limit reached; increase > kern.maxclusters > May 23 10:00:27 shiva2 last message repeated 2 times > > netstat -m shows this now: > > -bash-3.1$ netstat -m > 199 mbufs in use: > 195 mbufs allocated to data > 1 mbuf allocated to packet headers > 3 mbufs allocated to socket names and addresses > 194/6152/6144 mbuf clusters in use (current/peak/max) > 13860 Kbytes allocated to network (3% in use) > 0 requests for memory denied > 0 requests for memory delayed > 0 calls to protocol drain routines > > Can anyone offer any advice?
Double the mclpool limit and see if that is enough else double again. Stop doubling when you grow beyond 50'000 -- now it is time to reconsider and see if you have enough memory and kvm left for more. In general busy servers with many open connections need more then the 6000 clusters to work well. -- :wq Claudio