We're in the middle of a major CyberArk PAM overhaul and the high-availability design is giving us a headache. Our old setup was a simple warm standby, but it feels like a relic now.
I'm pushing for an active-active cluster for the PVWA and CPM components to handle our regional teams better, but the team is wary of the complexity. Is anyone running a true active-active setup in production? I'm especially curious about how you handle the database tier—did you stick with the vendor's recommendation or roll your own Always On / failover cluster? The docs aren't exactly clear on real-world trade-offs.
Active-active for PVWA and CPM is absolutely viable, but the database is your primary constraint. We bypassed the vendor's default SQL Server Always On recommendation after our own benchmarks showed unacceptable latency spikes during failover for our token-heavy workflows.
We implemented a multi-master PostgreSQL cluster with synchronous replication for the Vault database, treating it as a separate, critical tier. This required significant custom scripting for health checks and failover logic, but it eliminated the 8-12 second transaction freeze we observed with the warm standby model. The complexity is front-loaded; once the data layer is stable, the application tier clustering becomes straightforward.
Your team's wariness is valid for the initial implementation, but the operational burden shifts from manual failover procedures to automated state management. The real trade-off isn't complexity versus simplicity, but rather complexity now versus fragility later.
--perf
Wow, running your own Postgres cluster for this is pretty intense. I'm just starting to learn about PAM setups, so reading this makes me realize how much I don't know.
When you say "significant custom scripting," was that mostly for the failover logic? I'm curious how you handle monitoring to know when a switch is needed, or if it's fully automated now. That seems like the scariest part to get right.