Cumulus Support

Early this morning my one minute MySQL up-load failed. This has happened before. In the past (b3184 and before) this would quickly sort itself out and the upload would resume without my involvement. More recently a few have continued to fail until a restart.

This seems similar to the ftp upload problems I had in recent times that have been corrected with b3194 - see viewtopic.php?t=20525

Attached is the (edited) diag file. The last successful upload had a LogDateTime of 2022-07-17 00:34:00

In the early hours of the morning my ftp uploads, especially the realtime ones, often struggle to complete before the next cycle begins and this is often ... though not always ... the time the MySQL failures occurs

The log (5') and dayfile uploads happily continue despite the one minute's problems.

Thanks for the report. And yes there is a bug in there - the buffered failed commands are never removed from the list once they have executed OK so it keeps retrying the first one for ever.

Sorry to cause you problems but pleased you have worked out the cause. Also, the error doesn't trigger an alarm unlike the ones that stall then recover.

Because it was a bug the whole error handling was thrown out. Until this is fixed in the next release you could disable the "buffer on failure" feature if you have the odd issues with the MySQL connectivity. That way you may miss the odd update, but it should not get caught in this loop of always applying the same insert.

Thanks - I've disabled the buffer on failure

(post deleted, because quoted facts rejected)

Syntax errors ARE filtered out and not put in the failed buffer.

I see the user changing a table name as something a bit silly on do the fly, and a *very* edge case. You would hope that anyone changing their SQL schema would do so with CMX stopped - or at least not using it. So I will ignore this theoretical issue as I think it is unlikely - ha ha, never underestimate a humans ability to do the opposite of what you expect!

But you are correct, error conditions other than syntax could/should probably be checked, if you have suggestions as to errors that should not be buffered please make them. But on the whole the feature I think is a good one. I'm testing the bug fix now and it is recovering correctly when I remove/restore access to my test database.

A very belated but big thank you to Mark for sorting out the MySQL upload issues - since b3202 CMX has faultlessly handled all error generated by my system.