Atomikos Forum

I have a continuous build server on which I run some automated test scripts against software that's using Atomikos and XA. I have a number of build agents and a single database. We count on these automated tests running cleanly from a virgin environment.

The upshot of this is that I can't guarantee that my .epoch file is preserved between any number of tests that run against the database.

So my question is how do I get Atomikos to start over as if it had never run before. If I delete the .epoch file, the system will often refuse to reregister the database datasource. How does it know that this datasource exists? Is it implemented by the resource itself or does Atomikos have it stored somewhere?

My data sources are ActiveMQ+Sybase w/ DataDirect drivers.

The .epoch file ensures uniquess of transaction IDs across restarts. Deleting it may lead to "duplicate Xid" errors when accessing the resources, but not more than that.

So, presumably this is because the resources are tracking the XID somewhere and deleting the tm.epoch file doesn't get rid of them. If I know where the resources -- say Sybase -- stores it's XID's however, I could clear them out of there as well and eliminate the possibility of duplicate XID's?

Yes, if the resource really forgets about them. Not for production use, though...