Mailing List Archive

MX960 - DPCE cards / CPU usage
Hello,

We're running a couple of MX960 in our datacenter, carrying a full
table in BGP and running full MPLS services, fast-reroute/RSVP/LDP
etc. There are two LSP's in full mesh with two other routers. (the
960's are not edge routers, so no peering/transit terminate on them).

Hardware as follows:

2x RE-S-2000-4096-S
2x DPCE-R-20GE-2XGE
2x MPC-3D-16XGE-SFPP

The RE's are on 16.2R2-S11 which is the last service release supported
on these RE's. I wasn't sure if 17.3R3 supported 32-bit, or would even
work at all on these RE's?

I'm seeing the CPU on the DPC's hitting 100% in spikes every 5-10
minutes or so, which we did not see on JUNOS 13.3. The most concerning
thing is that we're seeing some packet-loss on traffic transiting the
DPC's at the same time. 1 to 2% packet loss every 10 minutes or so
through the router. This of course does not occur on the MPC which
runs smoothly with no spikes. There's no more than 400Mbps transiting
the router.

> show pfe statistics traffic fpc 6 | match discard
Packet Forwarding Engine hardware discard statistics:
Timeout : 0
Normal discard : 217186
Extended discard : 0

What would you suggest here? Is this being caused by the JUNOS version
we're running?

Any tips appreciated.

Thanks
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp
Re: MX960 - DPCE cards / CPU usage [ In reply to ]
Combination of DPC and full route is usually not a good idea, considering
the size of the full route these days.

Best thing you can do is of course replacing them with MPC cards.

If not possible I think FIB Localization might do the trick for you - must
say I havn't tried it myself, just make sure the routers uplinks are on the
MPC before you try it

https://www.juniper.net/documentation/en_US/junos/topics/topic-map/fib-localization.html


Daniel

On Mon, Apr 20, 2020, 21:29 Dario Amaya <darioamaya82@gmail.com> wrote:

> Hello,
>
> We're running a couple of MX960 in our datacenter, carrying a full
> table in BGP and running full MPLS services, fast-reroute/RSVP/LDP
> etc. There are two LSP's in full mesh with two other routers. (the
> 960's are not edge routers, so no peering/transit terminate on them).
>
> Hardware as follows:
>
> 2x RE-S-2000-4096-S
> 2x DPCE-R-20GE-2XGE
> 2x MPC-3D-16XGE-SFPP
>
> The RE's are on 16.2R2-S11 which is the last service release supported
> on these RE's. I wasn't sure if 17.3R3 supported 32-bit, or would even
> work at all on these RE's?
>
> I'm seeing the CPU on the DPC's hitting 100% in spikes every 5-10
> minutes or so, which we did not see on JUNOS 13.3. The most concerning
> thing is that we're seeing some packet-loss on traffic transiting the
> DPC's at the same time. 1 to 2% packet loss every 10 minutes or so
> through the router. This of course does not occur on the MPC which
> runs smoothly with no spikes. There's no more than 400Mbps transiting
> the router.
>
> > show pfe statistics traffic fpc 6 | match discard
> Packet Forwarding Engine hardware discard statistics:
> Timeout : 0
> Normal discard : 217186
> Extended discard : 0
>
> What would you suggest here? Is this being caused by the JUNOS version
> we're running?
>
> Any tips appreciated.
>
> Thanks
> _______________________________________________
> juniper-nsp mailing list juniper-nsp@puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>
_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp