[c-nsp] IOS Recommendation | 7600/RSP720-3CXL

[Raymond, Steven]

> I'm looking for a recommendation for upgrade and perhaps some possible
> explanation of the code branches.
>
> I am currently on 12.2(33)SRB1. The role of the box is Internet border,
> two full feeds, 10 other bilateral peers. No MPLS.
>
> Do I stick with SRB (I believe it's up to SRB6 now) or jump up to SRD2a?
>
> Any personal experience from the trenches would be appreciated.

We had bad results with all of SRB.  SRC3 seemed better but had a bad BFD bug that TAC couldn't resolve.

Have found the least bugs in SRD1, but non-cisco bgp neighbors sometimes require the use of hidden command "neighbor x.x.x.x dont-capability-negotiate" or the session won't restore.


SRB issues we remember:

SRB1
CSCek71050
Symptoms: Compared to other Cisco IOS software releases, unusually high CPU usage may occur in the BGP router process on a Cisco 7600 series that runs Cisco IOS Release 12.2(33)SRB1.

from SRB2 - very nasty bug with SIP/6700 linecards. linecard would stop forwarding, only fix was to reset both cards CSCsl50569

sh mem causes primary supervisor to fail, sh mem sum does not. Occurs in SRB2 and SRB3, only fix is upgrade

When deleting multilink member via 'no controller bla' if the multilink-group config is still under the serial interface it will eventually cause the SPA to reload. May take up to 24 hrs, also, no new multilinks will work until SPA is reset
Similar bug is when adding a multilink configuration to serial interface with service-policy applied, no multilinks will enable and SPA will eventually reload

SRB3
Caused router to misreport sampled traffic drastically as the linecards stop sending netflow.  CSCsq14299

SRB4
BGP phantom announcements stops sending routes to peers.  BGP is up and fine but neighbors get no routes, can be seen in "show ip bgp neighbor" as "prefixes total 0".  Possibly related to CSCsm57494.

SRB5
Appears to have same netflow bug as SRB3.
Tech added removed static route pointing to serial interface and added it back which caused standby to reload. Still undetermined if that was root cause