OVHcloud Network Status

Current status
Legend
  • Operational
  • Degraded performance
  • Partial Outage
  • Major Outage
  • Under maintenance
FS#5609 — CSCtr36441
Incident Report for Network & Infrastructure
Resolved
Allelouja !

After 4 weeks of discussions and exchanges with the TAC od Cisco
the support of Cisco discovered a bug on the Nexus 7000 platform.
This bug bothers us a lot

CSCtr36441 Traffic dropped on F1 peer-link for one leg vpc after ISSU

It is not yet externally viewable. Once the release note has been
reviewed it will be published at the above link. The details of the bug
are below:

########################################################

Symptom:

When shutting down one leg of a vPC, packets redirected through the peer
link are dropped.

Conditions:

- ISSU from 5.1(2) to 5.1(3)
- peer-link on a high port-channel number (i.e., peer-link carried on
Po4096)
- All F1 modules
- shutdown/disable vPC on single chassis creating 'one leg' vPC

Workaround:

This issue is not seen when the vPC peer-link port-channel number is
lower than the port-channels carrying vPCs. Therefore, changing the
peer-link port-channel number (i.e., reconfiguring the peer-link from
Po4096 to Po1) would resolve the issue. This would also bounce the
entire vPC domain so it may not be a feasible solution for most
customers.

Reload of both N7k peers

########################################################

This bug will be resolved in 5.2(1) which has not yet been released but
is planned to be released this month. There are two ways to resolve the
issue:

1) ISSU to resolved code once released
2) Since the issue is specific to an ISSU from 5.1(2), a reload of each
N7k will resolve the current state.

Please let me know how you would like to proceed. As always, feel free
to call or email me directly with any questions or concerns.


Thanks for the new Cisco team that follow OVH since 1 mont.
Thanks to thei (moral) support , we continued the exchanges
to come finally to a solution. The bugs are so complicated
and sophisticated that we need 1 month to find the origin
of the problem.

We still have 11 bugs like this ...
Posted Jul 11, 2011 - 20:29 UTC