[c-nsp] malloc failures disabling dCEF on GSR

[Adam Chappell]

Does anyone on this list have practical experiences of dealing with
low memory situations on a GSR?

Fragmentation eventually gets the better of this box and when malloc
fails, DCEF on the line cards is disabled in sympathy.

The net result is the GSR just sitting there not able to pass or
receive any IP traffic until I can get a field tech to power cycle it.

Aug  3 08:01:23 -Process= "BGP Router", ipl= 0, pid= 169
Aug  3 08:01:23 -Traceback= 5011E380 50135F10 5057C880 5054A348
5055432C 505544D8 507E8934 507E9580 50E17854 507E9610 507AF594
Aug  3 08:01:23 %FIB-3-NOMEM: Malloc Failure, disabling DCEF
Aug  3 08:01:57 %SYS-2-MALLOCFAIL: Memory allocation of 65536 bytes
failed from 0x50236564, alignment 0
Aug  3 08:01:57 Pool: Processor  Free: 1660428  Cause: Memory fragmentation
Aug  3 08:01:57 Alternate Pool: None  Free: 0  Cause: No Alternate pool

I realise the box obviously needs more RAM but some sort of feature
where I can get the box to crash and reload if it cant malloc would at
least mean less downtime.  Is anyone aware of anything like this?

Thanks.

-- Adam.