[quagga-dev 3064] Re: BGPD blocking

alex at pilosoft.com alex at pilosoft.com
Fri Apr 1 00:31:11 BST 2005


On Thu, 31 Mar 2005, John Payne wrote:

> 
> On Mar 29, 2005, at 9:31 AM, alex at pilosoft.com wrote:
> 
> > You are partially right - it is blocking in zebra communications, but
> > that's not the root cause. Root cause is that until bgpd completes
> > dropping routes from a peer it will not respond to any keepalives.
> 
> Would this still occur if zebrad isn't running, only bgpd?
Less so. It would still single-thread bgpd until all routes are dropped. 
In my test, 94% of CPU was used by Zebra, 6% by bgpd - so I would 
*suppose* if dropping a peer with actual kernel routes installed via 
zebra takes a minute, it'd take about 7 seconds by bgpd itself...

-alex




More information about the Quagga-dev mailing list