On Sat, Sep 26, 2009 at 5:43 PM, Bind <b...@dci.ir> wrote: > Hello > I have SunFire V880 (2 cpu +4G Ram) and installed bind 9.6.1-P1 on solaris > 10. > but my cpu load is very high!(above 90% during the pick time) > bash-3.00# prstat -a > PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP > > 562 root 2517M 2498M cpu0 0 0 1503:30:2 95% named/5 > 2394 root 3808K 3168K cpu2 59 0 0:00:00 0.9% prstat/1 > > here are some relatde information: > rndc status > recursive clients: 841/9900/10000 > My input traffic which shown by MRTG is about "2.5" Mbps recieved > requests(udp 53) and is normal in our network behavior during the pick time. > my question is: > does this high cpu load relate to input requests and is normall or it relate > to something else?
Can you get the number of request per sec? I think you can get that from output of two "rndc stats". On modern hardware it should be able to handle several thousands reqs per sec easily. Also, does named only use cpu0? It should be able to use all available cpus. If not, you should be able to force it using "-n". For comparison purposes, you might want to try using x86 server for DNS server and see the results. From my experience it's a lot cheaper and more powerful (compared to sparc or ppc) when used with bind. YMMV though. -- Fajar _______________________________________________ bind-users mailing list bind-users@lists.isc.org https://lists.isc.org/mailman/listinfo/bind-users
