bgpd bug

2013-09-27 Thread Thomas Greer
Hi All I'm seeing high CPU usage with bgpd session engine, and this was knocking out all my routing. The only way to get routing back is to pill the bgpd and then start it again. sthen suggested I ktraced it and the output is below. This is on 5.3 however same issues experienced on 5.2. Different

Re: bgpd bug

2013-09-27 Thread Stuart Henderson
On 2013/09/27 10:09, Thomas Greer wrote: > Hi All > > I'm seeing high CPU usage with bgpd session engine, and this was knocking out > all my routing. The only way to get routing back is to pill the bgpd and then > start it again. sthen suggested I ktraced it and the output is below. To clarify fr

Re: bgpd bug

2013-09-27 Thread Mike Belopuhov
On 27 September 2013 12:21, Stuart Henderson wrote: > On 2013/09/27 10:09, Thomas Greer wrote: >> Hi All >> >> I'm seeing high CPU usage with bgpd session engine, and this was knocking out >> all my routing. The only way to get routing back is to pill the bgpd and then >> start it again. sthen sug

Re: bgpd bug

2013-09-27 Thread Thomas Greer
On 27/09/2013 11:30, "Mike Belopuhov" wrote: >On 27 September 2013 12:21, Stuart Henderson wrote: >> On 2013/09/27 10:09, Thomas Greer wrote: >>> Hi All >>> >>> I'm seeing high CPU usage with bgpd session engine, and this was >>>knocking out >>> all my routing. The only way to get routing back i

Re: bgpd bug

2013-09-27 Thread Mike Belopuhov
On 27 September 2013 12:42, Thomas Greer wrote: > > > On 27/09/2013 11:30, "Mike Belopuhov" wrote: > >>On 27 September 2013 12:21, Stuart Henderson wrote: >>> On 2013/09/27 10:09, Thomas Greer wrote: Hi All I'm seeing high CPU usage with bgpd session engine, and this was knock

Re: bgpd bug

2013-09-27 Thread Henning Brauer
* Mike Belopuhov [2013-09-27 13:25]: > but your fstat output is sufficient as well for now. 0x9 > is a unix stream socket. > > > _bgpdbgpd52559* unix stream 0x80ccf000 <-> > > 0x80ccfe00 that should be the SE<->RDE socketpair, or SE<->parent, but that's less like

Re: bgpd bug

2013-09-27 Thread Mike Belopuhov
On 27 September 2013 13:46, Henning Brauer wrote: > * Mike Belopuhov [2013-09-27 13:25]: >> but your fstat output is sufficient as well for now. 0x9 >> is a unix stream socket. >> >> > _bgpdbgpd52559* unix stream 0x80ccf000 <-> >> > 0x80ccfe00 > > that should be t

Re: bgpd bug

2013-09-27 Thread Thomas Greer
On 27/09/2013 12:46, "Henning Brauer" wrote: >* Mike Belopuhov [2013-09-27 13:25]: >> but your fstat output is sufficient as well for now. 0x9 >> is a unix stream socket. >> >> > _bgpdbgpd52559* unix stream 0x80ccf000 <-> >> > 0x80ccfe00 > >that should be the SE

Re: bgpd bug

2013-09-30 Thread Thomas Greer
On 27/09/2013 13:17, "Thomas Greer" wrote: > > > >On 27/09/2013 12:46, "Henning Brauer" wrote: > >>* Mike Belopuhov [2013-09-27 13:25]: >>> but your fstat output is sufficient as well for now. 0x9 >>> is a unix stream socket. >>> >>> > _bgpdbgpd52559* unix stream 0x80c

Re: bgpd bug

2013-10-09 Thread Stuart Henderson
I've been taking a look on the machine, the EAGAIN is not from the SE<>RDE socketpair, but the TCP socket, I think there was some confusion between ktrace from one run and fstat from another before. When pushing routes out to the neighbour things stop progressing: Neighbor AS