In version 9.4.0 I created an iRule that, when the LB selects a pool member, it checks to see if the node is “down”. If it’s in a down state, disable persistence and force the LB to select another member.
(I haven’t been able to test this yet)
when LB_SELECTED {
if { [LB::status] eq “down” } {
persist none
LB::reselect
}
}
Some background info:
When our Web Administrators need to do maintenance on a pool member, they are able to bring a pool member down by renaming a file that the health monitor is tied to. Thus, the node gets marked down and they then wait for the connections to hit 0 before they perform their maintenance. What we are encountering is that persistence allows users to get back onto the pool member even though it’s in a “down” state. Rather than having to manually force the pool member offline, I created the iRule above in an attempt to see if we can stop the persistent connections in a more automated versus manual manner.
Is this a good approach or s there another way to do this?
Here’s our current problem, we have another BigIP device running 9.1.2. LB::status is not available in that version, so I’m unable to get check the selected node’s state/status. Any ideas on how to address/approach this problem on version 9.1.2?
Thanks in advance!!