We saw some backlog growing while issuing filecoin/accept receipts. This is due to not have pagination in the query that goes through submmited pieces pending a receipt.
Looking at the CRON execution duration, we can conclude for the last 5 days that average duration goes under 300k ms (5min). Which means we were having CRON alive 5 min and then 5 minutes sleeping. While we do not implement pagination on the query this seems a quick win.
Also note that if the CRON function ends up sometimes timing out, it is not really problematic. Looking at the code we run, we just go through table and, for each entry go through receipt chain to see if we can update the status (which makes a new lambda run to issue receipt). Therefore, if we timeout and restart the job, we will keep going where we were
We saw some backlog growing while issuing
filecoin/accept
receipts. This is due to not have pagination in the query that goes through submmited pieces pending a receipt.Looking at the CRON execution duration, we can conclude for the last 5 days that average duration goes under 300k ms (5min). Which means we were having CRON alive 5 min and then 5 minutes sleeping. While we do not implement pagination on the query this seems a quick win.
Also note that if the CRON function ends up sometimes timing out, it is not really problematic. Looking at the code we run, we just go through table and, for each entry go through receipt chain to see if we can update the status (which makes a new lambda run to issue receipt). Therefore, if we timeout and restart the job, we will keep going where we were