Terracotta Discussion Forums (LEGACY READ-ONLY ARCHIVE)

> Never run clustering on separate machines, unless their clocks are synchronized using some form of time-sync service (daemon) that runs very regularly (the clocks must be within a second of each other).

That is a scary warning.

What exactly can happen when the clocks are not completely synched?

And does this really only affect a cluster? How about an RMI server with clients that schedule jobs with time stamps that are off?

Follow-up question:

How about a cluster where I put all instances except one into standby mode at all time?

Would clock drift still be an issue in this situation?

Answer: Data corruption will occur.

And yes, if you have multiple scheduler instances pointed at the same DB, but only one has been started, then you will be OK.

What is the reason for the corruption?

What I really want to know is, is it possible to use Quartz in an environment where you are unable to synchronize the clocks to such precision? (We cannot necessarily get our clients to do so.) Perhaps by setting clusterCheckInInterval to a larger value?

The most likely problem with non-synced clocks would be false detection of failed nodes.

Setting a large check-in interval would help mitigate the problem - with the trade-off being a longer amount of time until a failed-node is detected and the job executions fail-over.