Just upgraded a cluster to 12.2.1, chasing those sweet, sweet "security enhancements" in the release notes. Now, the passive unit takes what feels like an eternity to claim the VIP during a simulated failover. We're talking 45-60 seconds of dead air, where previously it was under 10. This isn't a "brief interruption," it's a full-blown outage for any stateful connection.
Naturally, support's first response was to check our timeouts and carrier settings, as if we've never configured a cluster before. Everything is identical to the pre-upgrade config. The only variable is the firmware.
I've seen a few murmurs in other forums about changes to the session sync engine being more "thorough." Thorough is great for data integrity, terrible for recovery time objectives. My suspicion is the sync logic now waits for a complete handshake or validation of all session states before flipping the traffic, instead of the more aggressive, "good enough" takeover we had before.
Has anyone else been burned by this? Specifically:
* Noticed increased failover times post-12.2?
* Found any specific knob in the new firmware to revert to the faster, less cautious behavior?
* Or are we just expected to accept that High Availability now means "Highly Available... Eventually"?
prove it to me
Yeah, it's the session sync. They changed the default sync-delay-timer in 12.2. It now waits for session-table consistency before promoting the passive unit, which is why your failover looks clean in the logs but feels dead for a minute.
You can tune it, but it's buried. Look for the session-sync-hold-timer CLI command. Reducing it will get you closer to the old behavior, but you're trading consistency for speed. It's a garbage change for any environment with real RTOs.
Support's scripted response is useless because the default changed silently. The release notes call it "improved HA reliability" without mentioning the performance cliff.
Beep boop. Show me the data.