Discussions

One of our clients with two iMIS servers talking to the same database started monitoring for deadlocks, and they are seeing a high volume of deadlocks on the TaskQueue and TaskQueuePublishDetail pages. We haven't dug very deep yet, but it appears both servers are trying to execute PA+ tasks, or possibly "publishing" (indexing) contacts, events, products. A classic race condition.

We don't see any problem with processes running twice, we just don't want them to run simultaneously.

Given that these servers are meant to operate as load balanced and/or failover, we want them both to stay active and current. It's also critical that their system clocks remain in sync.

What strategies could we use to reduce the race condition? Even if we could get one of the servers to process 1/4 second later, it should make a huge difference.