F5 advanced troubleshooting

Say you have a version 9.x F5 Bigip. Nice loadbalancer, does some cool stuff. You add some new config and suddenly you have what looks like a layer 2 problem. So you console the box, and run top and the CPU (and maybe RAM) is pegged. Something called tmm is pegging the box.


TMM is a daemon that does most of the traffic management on F5. It is written by F5 and is opaque... except for:

tmstat

/usr/bin/tmstat from F5 gives you some output about what tmm is doing. So you can kind of figure out where things are going wrong, while you are dialing support.

Output looks vaguely top like:
NAME
CPU: 0% busy 1% idle 99% sleep Thu Apr 17 13:59:00 2008

Memory Allocated New Flow Old Flow Poll
21,081,060 / 1,807,745,024 99,851 25,284 1,228,703 Cycles
[ . : . | . : . ] 1 20 2,035 Total
vnic
Tc4,240b rx Cryplinkps 21,664b txass 27 Timers
[ . : . | . : . ] i8254x:00 (t[ . : . | . : . ]otal) 0 Stats
23,832b rx 1,000 linksa 4,112b txeudo
[ . : . | . : . ]ects 0 fu[ . : . | . : . ]tropy Virtual Class
Wa0b rx 0 linkecord 0b txcure 10,867,070 (total)
[ . : . | . : . ] 0 ci[ . : . | . : . ] 10,485,780 mco db
0 De0b rx 0 linkunseen) 0b tx 168,855 ssl
[ . : . | . : . ] [ . : . | . : . ] 143,727 tcl
0b rx 0 link 0b tx 68,708 (unseen)
[ . : . | . : . ] [ . : . | . : . ]
0b rx 0 link 0b tx Umem Class
[ . : . | . : . ] [ . : . | . : . ] 854 (total)
0b rx 0 link 0b tx 513 xfrag
[ . : . | . : . ] [ . : . | . : . ] 127 connflow
0b rx 0 link 0b tx 117 listener
[ . : . | . : . ] [ . : . | . : . ] 19 poolmbr
0b rx 0 link 0b tx 78 (unseen)