I’m planning to run two Temporal clusters, each in a different AWS regions, and I’m looking into XDC replication. The use case is long-running workflows which will occasionally be signaled and need to run a time-sensitive task. Without any sort of failover, if the cluster running the workflow is down, the tasks may not be executed on time.
I’ve read through the docs (Multi-cluster Replication | Temporal Documentation), but I don’t think I fully understand the limitations and tradeoffs that come with using an XDC setup. Can y’all help me answer these questions?
- Is there a recommended solution for this use case other than XDC replication?
- The docs specify “… replicates Workflow Executions from active cluster to other passive clusters.” Can Temporal XDC support an active-active deployment, where if one active cluster goes down, all of the workflows on the “down cluster” could failover and resume in the other cluster?
- The docs specify “data across clusters is not strongly consistent.” I imagine this means workflow tasks would be run at-least-once, instead of at-most-once. What are other symptoms of this eventual consistency (other than the version and execution history conflicts, which Temporal seems to resolve on its own)?
Thanks in advance!