bird-1.5.0-1.x86_64 crasches quite often on reconfiguration including renaming BGP protocols due to bad set of CFLAGS passed on compilation

Bartek Radwan groups at sq9mev.info
Wed Mar 30 09:26:36 CEST 2016


On 28.03.2016 12:21, Tomasz Pala wrote:
>> it's rather hard to reproduce such enviroment on my desktop.
> Why? On desktop - use private ASn, on BGP router - use tables not


Well, i did not mention i have two bird instances, ant this "choking" is 
observed on just one of them - the one which has more bgp session.


bgp-1:
# birdc show protocols | grep Established -c
14

# birdc show route count
1612660 of 1612660 routes for 581641 networks


bgp-2:
# birdc show protocols | grep Established -c
3
1261789 of 1261789 routes for 581643 networks

The issue is observed only on bgp-1 (the one with 14 established sessions).

Simulating 14 sessions is quite easy in fact, and i'll try to do that, 
however i'm not sure if the primary issue is just number o estabished 
sessions, or process activity triggered by bgp messages received by some 
of them.


Since last post i've installed 1.5.0-3 (my fork with backported patches 
from upstream), but the issue still remains. Then i've reverted do 
1.5.0-1 available in th now and loadavg is stable now.

What i've noticed is that average value of loadavg is in both situatinos 
the same value - about 0.3. With 1.5.0-[2,3] reaches quite low and quite 
high values, while with 1.5.0-1 it's more constant near 0.3

http://www.sq9mev.info/upload_29_3_2016_at_09_17_46.png

This observation is so valuable that I can quickly determine if the 
problem persists, without waiting for the peak of loadavg.

I'll post the group when i find out more.

I'll try to remove -fwrapv.

-- 
Regards
Bartel


More information about the pld-devel-en mailing list