In T343952#9124449, @KCVelaga_WMF wrote:

@DamianZaremba

I was looking a bit further into the percentage of the reverts made by ClueBot NG that were reverted back. Although all of these are not necessarily false positives, it's interesting to see that the percentage of reverts reverted back has come down by about 50% over time (~16% at the beginning to ~8% in June 2023). I am wondering if you have any insight into what's causing this (it can be that the model is re-learning consistently, or the decrease in overall revert rate might be a factor, or something else).

In terms of the bot, there's been no notable configuration changes made over the described time period; quite literally the configuration has not changed from the currently deployed version in 2 years (previously was the same, but 2 years ago was the last build in CI and is the as-deploy artefacts x-ref https://github.com/cluebotng/utilities/blob/main/fabfile.py / https://github.com/cluebotng/core/releases/tag/v1.0.2).

After some debugging I figured out what is happening here.

I will have a look later, but I suspect this is due to a stuck task which has seemingly also been affecting restarts in the last weeks; unless the schema grew drastically in the last months. These dumps are primarily used for development and ad-hoc copies of the history for archival, so the current 1 week retention can be reduced if required.

In T293611#7435665, @Reedy wrote:

I know you said "all options selected", but just to confirm "High-volume editing" is definitely checked?

I deployed https://github.com/cluebotng/report/releases/tag/v1.0.3 which should sort it.

In the past 2 days the bot got killed 5 times for excess flood.

This was turned off for a while, but apparently got turned back on again without actually being fixed.

The bot is configured to identify and get voice to avoid this issue, but it appears that at some point it timed out and on re-connect didn't, so ended up not voiced.

This is because the core was broken (due to debugging something else), so the bot had no score data; in that case the score is null or as far as IRC is concerned an empty string. This is expected as anything post-core is 0-1.

The format hasn't changed AFAIK, comments are not escaped, the format is as below:

https://meta.wikimedia.org/wiki/Steward_requests/Miscellaneous#OAuth_consumer_for_ClueBot_NG

I've updated the crontabs to restart the processes with no stdout/stderr logging enabled (https://github.com/DamianZaremba/cluebotng/commit/3d026466688691972563313a6f1f8720c3efe714) to ensure the /dev/null doesn't get reverted and re-deployed the code.

I've cleaned up the code in both bots and put in MonoLog with a daily rotation. I'll keep an eye on the size for a couple of days to make sure things are looking ok.

Thanks @chasemp! I can see the logs are current linked to /dev/null.

I replied to the mailing list about this, in reply to the email about NFS running out of space.