Error about non-consecutive data in history

[Stefan Ritt]

Once we had a crash of the logger, and afterwards the history (FILE) was corrupted. This lead to tons of error message

Detected duplicated or non-monotonous data in history…

Yes we know now that there is a problem in the history database, but seeing this error about 10’000 times every day obscures some real problems. For the moment I just disabled this message, but there a certainly better solutions: If the error occurs, it can be silenced for some period (hour, day), or the error could be suppressed by some ODB setting. Alternatively, a program to fix the corrupted history system would also be a solution.

[dd1]

there was a bug in mlogger that caused writing of invalid history files, now fixed. (I forget the details, it is probably posted to the midas forum).

the files that produce these errors are broken and should be removed/repaired.

if somebody provides me with an example broken file, I can write a repair tool.

history reader uses binary search and this requires that time monotonically increase across the history data file.

K.O.

‌

[dd1]

error message restored. it is important, history will not work correctly in the presence of broken history files. K.O.

[Stefan Ritt]

I strongly disagree to that. We had the case where we had some wrong event in our history file. It’s important to know that, but it’s annoying to see this error message on every history access for about a year. If this error is active, we desperately need a tool to repair the history in that case. Getting one error message about every second in a large collaboration clutters up the messages and makes real errors almost invisible.

Before we have a repair tool, I request to disable the error, otherwise we cannot run at PSI.

[dd1]

from private communication with stefan, he is okey with closing this bug report. K.O.