Slow Spool Enumeration in KumoMTA

I’m encountering a performance issue with my KumoMTA server. The logs are showing that the spool process is taking a significant amount of time to enumerate items. Over time, the enumeration rate improves, but it’s still quite slow.

Here’s a snapshot of the logs:

Aug 24 05:57:18: still enumerating. 0 items in 120s
Aug 24 05:57:48: still enumerating. 0 items in 150s
Aug 24 05:58:18: still enumerating. 0 items in 180s
Aug 24 05:58:48: still enumerating. 1348 items in 210s (7 items/s)
Aug 24 05:59:19: still enumerating. 18111 items in 240s (76 items/s)
Aug 24 05:59:49: still enumerating. 29789 items in 270s (111 items/s)
...
Aug 24 06:05:49: still enumerating. 140736 items in 630s (224 items/s)

It seems that initially, the server is not processing any items, and then gradually starts handling more as time progresses.

I’m concerned that this might indicate a deeper issue with how the spool is being processed or how resources are being managed. Has anyone experienced similar behavior or have insights into what might be causing this?

Please read and do the above :backhand_index_pointing_up:

This is likely related to that full disk situation in your other post.

Yeah, that makes sense. The full disk issue is probably the root cause here. I’m working on figuring out how to safely clear out space and manage the disk usage better moving forward. Any tips on handling the spool and logs to avoid this in the future would be super helpful. Thanks for connecting the dots!

The newer dev version has a disk usage limit to prevent it by providing temp fail to injectors.

But you’ll still need to monitor log space and retire old logs.