Forum Discussion
Mikhail_Smirnov
Nimbostratus
NimbostratusZebOS routing metric control from LTM
Hello guys!
Our infrastructure is a network running IP protocol and several datacenters in it. The network uses OSPF (may be it will be IS-IS in future) for dynamic routing. Each datacenter has its LTM and server farm serving identical virtual servers with identical IP address. High availability is made by ZebOS on routing level. The idea is use ZebOS to make virtual server not only high available but also make load balancing between datacenters depending on number of alive members in a pool in each datacenter. To do that we should be able not only do Route Health Injection but we must have a hook to adjust routing metric in each ZebOS according to number of alive pool members in each virtual server.
Any ideas? I couldn't find something similar here.
Cheers.
2 Replies
mazek_59373Hi,
have a look at GTM, afair there is a load balancing method based on available pool members. You could then spread the load on vservers based on live members in pools.
cheers
maz- it is easy to get RHI to be off if a minimum number of pool members are not present because you'd actually be disabling the virtual server and RHI would withdraw the route.
( you do this by monitoring an virtual server with an irule that checks `if { [active_members] >= X }` where X is the minimum number of pool members.)
To manipulate the routing metric itself, I think you would have to fall back to a script running on the Linux host that periodically checks the number of active members ( b pool external_gateway_pool show |grep active,up |sed -n '$=' ) and then modifies the routing metric using the zebos vtysh, which looks like it would require an Expect-like interaction.
You have bash, tcl, perl and python (in 10.1) to pick from when writing the script.
