Forum Discussion
NimbostratusIs there a GLB back off timer for failed pools
Hi, I've been searching and testing this for a couple of days but not been able to work it out, wondering if anyone knows.....?
I am testing GLB with topology based LB. There are 2 pools and all appears to be working fine (currently it is in testing, not live service). If I fail pool1 then all traffic which the topology records would steer that way automatically go to pool2. The same goes for pool2 failing.
When I re-introduce the failed pool traffic does not go to it for X period of time. It's like it's doing a slow start. Can anyone point me towards an article or let me know how the DNS module handles failed pools coming live again?
Many thanks
BaselineI think I may have just stumbled across my own answer. If I look in DNS > GSLB > Wide IPs and then choose the Wide IP I'm interested. Under Pools it has Persistence = Yes with a TTL of 3600.
My assumption is that it will keep all traffic going to the second pool until the TTL timer expires? Presumably the TTL timer resets to 3600 each time it receives traffic so the client will have to send no traffic to the GTM for 1 hour before it will start using its preferred pool again.
If so, is the client the end user, or the user's local DNS server? If it's the local DNS server, in our scenario (we're a 24x7 business) the timer will rarely get to 0. Therefore, it's probably best for persistence to be turned off, or lowered to 1 or 2 mins?
- Baseline
As per my comment, I appear to have stumbled across my own answer!
Setting the persistence to off allows the clients to immediately revert back to their preferred pool when it becomes available again.
Andrew