[Devel] 100% cpu
helge at smartnet.no
Wed Oct 25 11:13:32 CEST 2006
after some random testing I found my problem.
I received a 484 address incomplete from my PSTN provider when a
customer dialled a non-existant PSTN number (not an 404 which u usually
get when setting up a local PSTN gateway), which created a CANCEL loop
in my cluster configuration. I'm not sure if this is a UAC or a UAS
problem, but at least I have something to work with.
man, 25,.09.2006 kl. 17.21 +0300, skrev Bogdan-Andrei Iancu:
> Hi Helge,
> looks like a memory problem - fm_status is not able to do proper memory
> listing due some corruptions - try to compile the QM_DBG_MALLOC support
> and run the proxy - eventually the memory manager will report a bug
> Helge Waastad wrote:
> >I suddenly got 100% cpu usage on my openser processes..
> >I tried to kill the daemon and heres the backtrace of the remaining
> >0x0000000000456790 in fm_status (qm=0x5eca60) at mem/f_malloc.c:515
> >515 for (f=qm->free_hash[h].first,j=0; f;
> >(gdb) bt
> >#0 0x0000000000456790 in fm_status (qm=0x5eca60) at mem/f_malloc.c:515
> >#1 0x000000000041f172 in sig_usr (signo=Variable "signo" is not
> >) at main.c:571
> >#2 <signal handler called>
> >#3 0x0000003a57ac72ee in __recvfrom_nocancel ()
> >from /lib64/tls/libc.so.6
> >#4 0x000000000044f2d6 in udp_rcv_loop () at udp_server.c:415
> >#5 0x000000000041f7f4 in main_loop () at main.c:925
> >#6 0x0000000000420446 in main (argc=Variable "argc" is not available.
> >) at main.c:1477
> >I do not know why this happened, but the system consists of to servers;
> >both running an dispatcher and a sip router (and a sql cluster node)
> >And both servers went bananas.
> >Any clues?
> >br hw
More information about the Devel