On Sun, Sep 14, 2008 at 02:57:19PM +0200, marcin wrote: > Hi, > > I have network server running on OpenBSD 4.3. Kernel was patched and > recompiled with changed amount of queue for hfsc and cbq to 768. > This machine has 3 interface, two external and one internal. > On each interface i have hfsc queing active. On server i have DHCPD server, > DNS (bind), PF+queing, Postfix + sasl + postgrey + clamd for small group > of users. > Server working for 100 hosts. > > My problem: In my logs i'm getting following message: > dhcpd: send_packet: No buffer space available >
Most porbably it has todo with your altq setup. If the packet can't be queued because the bandwidth or queue length is exeeded it will be dropped and a ENOBUFS is returned. > I tried to change my NIC, change values for sysctl, but this did not help > me... Someone had similar problem ? > Below I included additional information which can help to set the cause > of problem > > Thank You! > > # sysctl > > net.inet.udp.sendspace=262144 > net.inet.udp.recvspace=262144 > Does not matter but honestly you should not fiddle with these settings if you don't know the effect of the values. > net.bpf.bufsize=10485760 > net.bpf.maxbufsize=10485760 > I doubt this is the limiting factor in your case and I don't want to know in how many ways setting the bpf buffer size to 10MB will cause troubles later on. -- :wq Claudio
