[j-nsp] ldp flaps after bgp reconfiguration ?

[Alexandre Snarskii]

On Fri, Jun 17, 2005 at 10:20:37AM -0700, Pedro Roque Marques wrote:
> Alexandre Snarskii writes:
> 
> > Any comments or suggestions on that problem ?
> 
> A bit too little information.

Sorry about that, i'm a bit novice in junipers :( 

> Do you see any message in the logs about "scheduler slips" ? You can
> try enabling task accounting (on cli "set task accounting on") which
> will give you more detailed logging.

Yes, we do see these messages: 

Jun 20 07:17:21 lo0.ProblemRouter rpd[2706]: task_reconfigure reinitializing
done                                                                        
Jun 20 07:17:22 lo0.ProblemRouter mgd[67655]: UI_COMMIT_PROGRESS: commit:   
signaling 'SNMP daemon', pid 2704, signal 31, status 0                      
Jun 20 07:17:39 lo0.ProblemRouter rpd[2706]: RPD_SCHED_SLIP: 9 sec scheduler
slip                                                                        
Jun 20 07:17:49 lo0.ProblemRouter rpd[2706]: RPD_SCHED_SLIP: 8 sec scheduler
slip                                                                        
Jun 20 07:18:10 lo0.ProblemRouter rpd[2706]: RPD_SCHED_SLIP: 13 sec         
scheduler slip                                                              
Jun 20 07:18:10 lo0.ProblemRouter rpd[2706]: RPD_LDP_SESSIONDOWN: LDP       
session <PeerRouter> is down, reason: received notification from peer         

> 
> >  Should we modify logic of our script to perform one commit per
> > modified as-path-group ?
> 
> There are situations where the task of reading the config database
> itself (not really processing it) can take several seconds... in this
> case you would see a message about a scheduler slip while processing
> config. If that is the case, do you have config groups ?

No. 

> How does your script update the config ? e.g. "load replace" / "load
> override" ? 

load replace

> do you replace all the policies ?

Not, just the modified ones. (And we do not see exact correllation between 
number of updated policies and did the sessions dropped after). 

> > Or should we just upgrade junos to more modern version ?
> 
> There have been optimizations/fixes in recent releases that address
> problems w/ similar symptoms.

As one of our engineers said to me - we saw that problems every 
day before we upgraded to 6.4R3.4. With that version - just some 
per month.