PO Suggestions Taking Forever

Have an odd one that I’m stumped on, hoping someone can assist.

We are multi-company, on-premise. We run our PO Suggestions at midnight, noon, and 7pm. Everything had been fine for years, until last week. Our main company went from taking 30-40 minutes to run to now taking 4 to 5 hours.

These run in a scheduled process set, one run for each company. It started last week, becoming gradually worse until current state. We didn’t catch it last week as it only took about an hour, then Monday came and it took 2 hours, I brushed it off, big mistake. Tuesday it took even longer, that’s when we decided to turn off the noon and 7pm run, as they would overlap. This morning, it took nearly 5 hours to run our biggest company, and 2 of the smaller company’s times have doubled. The other companies seem to be okay.

This is causing major issues for our purchasing guys, they would normally process suggestions in the morning, then process any new ones that came in after lunch (noon run). They are not getting the noon runs. The PO suggestions process also prevents sales from using CTP while it’s running.

I’m being hounded from multiple directions to get this fixed now. I’ve reached out to Epicor Support, provided logs, and they haven’t been much help. The runs do complete without issues, so there’s no “Errors” in the logs themselves, it’s just taking way longer than it should, like a second per part, maybe 2 parts per second, when it should be running several parts, or more, per second.

Our resources and resource groups have not changed. I’ve tried the “recalculate part low level code” process as well, it didn’t seem to help.

Any help is appreciated!

Hosted onprem setup here. Could depend on a lot of things…load on the app and/or SQL server at the time (we have separate boxes)…things OUTSIDE of Kinetic (backups, Windows updates, antivirus updates).

MRP and PO suggestions run early AM via system agent…MRP normally done 15-25m range…PO suggestions net change ~5, regen ~25.

Earlier this week we had a random case where the MRP was still running after 4 1/2 hours and was only on item number “00872” …really early on in the cycle (assuming it’s running purely in item number sequence). Task had to be killed off, reran on scheduler the next night and was back to the normal runtime.

Sounds similar to us. We also have separate boxes for App and SQL. I will check backups, I do plan on rebooting the servers tonight and running any updates they might have. Antivirus is worth checking, even though we’ve had the same protection for the last year. Thanks for the tips.

How many processers are you running? If you look at your logs each process will have its own plus an unnumbered one.
If you look in the log is any one part or group of parts taking a long time to calc?
Are you running mrp recalc before suggestions in the process set?

Have you dumped PartDtl to see if there are any bad records?

No problem…we’re at the mercy of our hosting company but they’ve generally been pretty good with their update/patch processes.

Except for that one time where nothing ran that day…something about Crowdstrike Falcon or Eagle or Buzzard…

1 Like

Are sql statistics being recalculated in a maintenance routine on a regular basis?

Assuming this would be found under maintenance plans in ssms? This is what I see. I’m guessing no, we’re not, and I’m not sure how to set that up. Something google can teach me?

image

I did enable “Query Store” the other day after looking into this. Someone recommended forcing a plan… Problem is, we didn’t have query store enabled before the issue so I have no known good runs to force a plan from.

Epicor recommends these and I would bet most of us here use them. They are easy to setup.

This thread may have some hints. I would do the update to start.

ALTER INDEX ALL ON erp.PartDtl REBUILD
GO
UPDATE STATISTICS erp.PartDtl WITH FULLSCAN 
GO

https://www.epiusers.help/t/mrp-delete-step-takes-forever/94629/4

1 Like

Awesome, I’ll check them out and give it a go after my reboot and updates tonight.

Ended up doing a server reboot last week and installed pending updates. That was the magic bullet for us.

2 Likes