Recently (on 3 July 2024), the bot I had running on Deucalion was reacting multiple times to the same event, resulting in a series of identical comments, see for example EESSI/software-layer#630 (comment).
@Neves-P mentioned that he saw a similar problem with a test instance of the bot he was playing with, in a totally different context (at RUG), so it seems like a general problem, not specific to the setup I did on Deucalion (which had been working fine for days already before this happened).
To me, it seems that smee.io is actually the culprit here, and that it was echo'ing events coming from GitHub multiple times to the smee channel used by the bot.
That would also explain why the production bots running in AWS & Azure were not affected, because they're using a custom smee instance (smee.nessi.no).
We should:
- try to figure out exactly what happened;
- see if we can somehow make the bot robust against this, for example by letting it refuse to re-process an event that has already been processed (but that means keeping some state, I guess keeping the last 100 event IDs would be sufficient in practice);
Recently (on 3 July 2024), the bot I had running on Deucalion was reacting multiple times to the same event, resulting in a series of identical comments, see for example EESSI/software-layer#630 (comment).
@Neves-P mentioned that he saw a similar problem with a test instance of the bot he was playing with, in a totally different context (at RUG), so it seems like a general problem, not specific to the setup I did on Deucalion (which had been working fine for days already before this happened).
To me, it seems that
smee.iois actually the culprit here, and that it was echo'ing events coming from GitHub multiple times to the smee channel used by the bot.That would also explain why the production bots running in AWS & Azure were not affected, because they're using a custom smee instance (
smee.nessi.no).We should: