Friday, December 7, 2018

Problem with Juniper ACX during link failover in MPLS core

Here is a simplified architectural diagram

We operate an ISP style backbone (comprised of Juniper ACX1100) where we run IS-IS over every link. We then run LDP over the links for label distribution

We also run a BGP-free core where the PE routers are BGP peer'd backed to the Core and we have a VRF for transporting Internet traffic.

I am having a peculiar issue with a particular router, P2. When we disable the link (as depicted in the diagram), P2 cuts over to the working path (P2 -> P1 -> Core1)

The Internet (VRF) circuits at PE1 and PE2 continue to work perfectly fine (as expected). However, the VRF in P2 does not work properly, not forwarding Internet traffic correctly to the Core. Checking the logs on P2, I see a few of the below messages:

Dec 4 13:41:09 [redacted] feb0 ACX_NH::acx_nh_tag_hw_uninstall(),2326:acx_nh_tag_hw_uninstall: nh 815 egress uninstall failed: (-10:Operation still running)

Dec 4 13:41:09 [redacted] feb0 ACX_NH::acx_nh_ucast_uninstall(),2452:acx_nh_ucast_uninstall: tag uninstall failed, err: -10

It seems like the ACX is not able to remove the old next-hop from the forwarding table. Has anyone experienced this before? I am going to schedule a maintenance window to reboot the unit to see if that can be the problem - but I would like to first understand why ths happening so that we can avoid future issues.

I am running JunOS 15.1R6.7



No comments:

Post a Comment