[c-nsp] 12.2(33)SXI revisited

A.L.M.Buxey at lboro.ac.uk A.L.M.Buxey at lboro.ac.uk
Fri Feb 20 15:35:31 EST 2009


hi,

just thought I'd post an update.... it seems that we hit 3(!)
bugs in the upgrade process. 2 of them are now documented
and one is in the process of being documented.

fundamentally, SXI has a new default line added to the
config...which meant that our second supervisor couldnt
come up....we fixed this by removing the offending line
in the working sup720 - but by this time, it was too
late - the system thought there was a failure and
then attempted to use ISSU to recover the previous image.
ah..oh. we upgraded from SXF15a so there wasnt
ISSU activity....(this also affects upgrades from SXH).
anyway, this is where the oif and ltl came from - what
had now happened is that any multicast groups that were
being joined got their IGMP join locked in the MLS
switching fabric...stuck.... never to work. only groups
that hadnt been borked during the dual sup720s being
present would work.  

the instant fix was to switch off hardware switching of
multicast to flush these stuck registers etc and then
very quickly(! ;-) ) turn it back on. 

we await resolution of the dual 720 bug that we hit..but
a reload is needed as part of that..so has to wait a little
longer.

thanks to Cisco engineers involved in looking at this case.
it was certainly 'interesting'

alan


More information about the cisco-nsp mailing list