|
|
On startup, cnxmq (EP) reads all .q queue files (_mdt_??? data files) and, when it detects a corrupted file, deletes it.
Instead of deleting, how about creating a directory (e.g. named corrupted.<PID>) and moving them there, so they can be examined later to determine the cause of the crash, and possibly even recovered one day in the future?
In the case of EP (but not cnxmq) I think you're already doing something similar by packing them into a file ep_crash.zip. The problem is that it's always the same filename, and usually after a "real" crash that causes corruption, EP crashes again, this time with empty data files, and the empty ep_crash.zip overwrites the interesting one so it's lost (so how about calling it ep_crash.<PID>.zip?)
| Use Case | Replication |