[afnog] CFEB issues on juniper

[ibtisam jamal]

How can i check for this  "failure of the software
to delete inactive routes from the FIB" ?

see the log below ,we had to restart ...on router 2 .what does the log mean
.


7 13:34:32  SR02 cfeb NH: invalid nh (125291) for bulk statistics
Feb  7 13:34:44  SR02 /kernel: RT_PFE: NH IPC op 30 (GET NH BULK STATS)
failed, err 1 (Unknown) peer_class 0, peer_index 0 peer_type 1
Feb  7 13:34:44  SR02 /kernel: RT_PFE: NH IPC op 30 (GET NH BULK STATS)
failed, err 7 (Doesn't Exist) peer_class 0, peer_index 0 peer_type 1
Feb  7 13:34:44  SR02 /kernel: RT_PFE: NH IPC op 30 (GET NH BULK STATS)
failed, err 1 (Unknown) peer_class 0, peer_index 0 peer_type 1

the second router has the below log ........

Feb  8 11:23:01 SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:23:04  SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:23:07  SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:25:07  SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:25:29  SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:25:31  SR01 cfeb CPROD: daemon socket creation failed
Feb  8 11:25:35  SR01 cfeb CPROD: daemon socket creation failed



On Tue, Feb 7, 2012 at 6:40 PM, Mark Tinka <mtinka at globaltransit.net> wrote:

> On Tuesday, February 07, 2012 09:10:26 PM ibtisam jamal
> wrote:
>
> > We had a similar issue on one of the internet routers and
> >  the heap utilization would go up and we kept monitoring
> > and if it got to around 90% we restarted the router ,as
> > trial and error, upgraded to junos 10.4R5 then also
> > installed RE-850.
> >
> > The heap utilization has been at 62% ever since .
>
> 62% is a little high, but well below limits I would consider
> worrying. Just keep a close eye on it.
>
> > It is not clear what solved the issue because we also
> > installed new routers with RE-850 for internet traffic
> > only and the routes are as below the only difference is
> > the RE is RE-850..
>
> The RE-850 is purely a control and management plane
> component. All data plane services are handled by the CFEB,
> regardless of which RE is in use.
>
> The reason your other router is fine is because you were
> able to reduce its Heap utilization. Once the Heap
> utilization becomes excessive, you end up the issues you're
> seeing.
>
> What you need to investigate is why the Heap utilization
> reduced after a software upgrade + RE swap-out? It could
> also have been due to a reboot, or a failure of the software
> to delete inactive routes from the FIB prior to the upgrade
> (crazier things have been known to happen).
>
> This definitely sounds like one for JTAC.
>
> > We do not have RE-850 now and we do not have CFEB-e .
>
> If you keep holding these many routes in those routers,
> sooner or later, you'll need to upgrade these components.
>
> > I have also noticed this could it be a contributor to the
> > issue
> >
> > mpls.0: 124645 destinations, 124645 routes (289 active,
> > 124356 holddown, 0 hidden)
> > Restart Complete
> >                 MPLS:      3 routes,      3 active
> >                 RSVP: 124635 routes,    279 active
> >                  VPN:      7 routes,      7 active
>
> While your RSVP routing table has more routes than I'd
> normally see (are you importing a large portion of inet.0 or
> some other heavy routing table into RSVP?), you only have
> 279 routes actually installed in the FIB, so no major drama.
>
> Mark.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://afnog.org/pipermail/afnog/attachments/20120208/0eed137f/attachment.html>