Please fix the weekly digest error recovery issue
-
I believe there is already a discussion about this but the weekly digest fails for me 14 times out of 16. What's happening is that the emails are being sent too fast with too many errors and it all craps out. Not being able to send the weekly digest is costing ad revenue loss plus we're not realizing the incremental benefit of bring people back into the forum conversations, which is the more important thing.
- Run the digest in batches
- Use exponential retry logic
- Allow admin to manually resume where it left off
I think the last one is the most important because at the very least I can babysit the process weekly until more automatic logic is released. It's almost impossible to even debug or try to do anything myself because I don't want to re-send the first 100 - 500 successful messages.
Part of the issue is that I have to SparkPost, I'm on a legacy free plan and I get 100k monthly emails. The existing nodebb-spark-post mailer doesn't seem to work so I have to rely on Spark Post itself to handle not block previously bounced emails. I'm not willing to move to another provider because it would cost about $100/month.
I'm still on 1.11.x because the emailer worked last week but failed again this week. I didn't want to change anything. So far, since November 13, the weekly digest has worked twice.
Finally, I can't post what the error is because log rotation is too fast on NodeBB. I basically lose all of my logs on a rolling 10 hour basis.
-
Related Github issues: