[FROG] BGP CPU Usage

[Donald Sharp]

Richard -

We've been investigating and fixing performance issues.  A config
would help us to a small extent because there might be a more
efficient way to organize your peers.  But what what we would really
like is a flamegraph of bgp performance, to see what areas of
inefficiency you are hitting.

https://github.com/FRRouting/frr/wiki/Perf-Recording

donald


On Tue, Oct 17, 2017 at 7:52 AM, Richard J Palmer <richard at merula.net> wrote:
> Hi
>
> We have FRR running on a number of core routers here - However it seems that
> bgpd is using 100% CPU most of the time:
>
> top - 12:48:16 up 95 days, 20:02,  1 user,  load average: 1.41, 1.25, 1.20
> Tasks: 151 total,   4 running, 147 sleeping,   0 stopped,   0 zombie
> %Cpu(s): 11.7 us,  3.6 sy,  0.0 ni, 83.8 id,  0.0 wa,  0.0 hi,  0.9 si,  0.0
> st
> MiB Mem : 3937.914 total,  509.301 free, 2035.738 used, 1392.875 buff/cache
> MiB Swap: 3082.750 total, 3082.750 free,    0.000 used. 1410.402 avail Mem
>
>   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
> 23010 root      20   0 1651212 1.515g   3764 R  93.0 39.4   1594:12 bgpd
>  1050 root      20   0  409292 347104   3676 R  36.2  8.6  48720:02 zebra
>     7 root      20   0       0      0      0 S   0.3  0.0 866:46.42 rcu_sch
>
> There's only 15 peers mostly internal (one transit link). I am using both
> Ipv4 and IPv6
>
> I am happy to share the bgp config as needed - biut there's nothing obvious
> I can see ... And the log is not showing any substantial entries
>
> Is there something I am missing / or would give a hint on what is causing
> this high load / and how to resolve it ?
>
> Thanks
>
> Richard
>
>
> _______________________________________________
> frog mailing list
> frog at lists.frrouting.org
> https://lists.frrouting.org/listinfo/frog
>