Solved: CSM migration (interface routing)

stephendrkw · ‎09-20-2010

I've started replacing our CSM's with bigip F5's. I needed to add some routes (example below) so real servers are able to route back through F5 (10.138.100.15), before removing the CSM module. My question is.... I'm in the process of migrating a few servers every weekend. Is there any limitation for the route command, how many route statements on a CSM interface? I need to add route 10.x.x.x x.x.x.x gateway x.x.x.x for approx 500 real servers /32 hosts for 2 weeks, as I'm unable migrate a whole VLAN in one hit!

vlan 24 client

ip address 10.138.100.11 255.255.255.0 alt 10.138.24.12 255.255.255.0

gateway 10.138.100.1

route 10.138.30.0 255.255.254.0 gateway 10.138.100.15

route 10.138.27.0 255.255.255.0 gateway 10.138.100.15

route 10.138.25.0 255.255.255.0 gateway 10.138.100.15

Gilles Dufour · ‎09-20-2010

The CLI will restrict the static route configuration to
(4K - 1) entries on each VLAN.  

There is a hardware memory space of 1.5 Megabytes
to hold the routes lookup tree.  Each entry requires
2 bytes; So, the maximum is 0.75 M entries.  This tree
holds all kinds of route: gateways, routes, vlan, learned hosts.
It is a prefix tree...So it is hard to say whether the routes
are packed or not.

Too bad you're leaving us for F5 ...

I hope you will come back.

Gilles.

View solution in original post

Gilles Dufour · ‎09-20-2010

The CLI will restrict the static route configuration to
(4K - 1) entries on each VLAN.  

There is a hardware memory space of 1.5 Megabytes
to hold the routes lookup tree.  Each entry requires
2 bytes; So, the maximum is 0.75 M entries.  This tree
holds all kinds of route: gateways, routes, vlan, learned hosts.
It is a prefix tree...So it is hard to say whether the routes
are packed or not.

Too bad you're leaving us for F5 ...

I hope you will come back.

Gilles.

stephendrkw · ‎09-20-2010

Thanks Gilles, that leaves plenty of statements available.

(Why F5? Well at the time CSM's could only handle 240 continuous tcp health checks which wasn't enough for us, especially for trading systems where most health checks were 1-3 seconds. The CSM processor for the health checks had reached its limit and whenever a server was lost the CSM didn't have the ability to take the server out or service for sometimes 2 mins or more if I remember. Example below......

#sh mod csm 3 tech-support probe
Software version: 4.2(6)

--------------------------------------------------------------
------------------ Health Monitor Statistics -----------------
--------------------------------------------------------------
Probe templates: 17
Suspects created: 1477
Open Sockets in System : 240 / 240
Active Suspect(no ICMP): 10 / 200
Active Script Suspect : 0 / 50

Anyway I understand the ACE handles over 1,000? Maybe we might switch back to ACE in the future!