Sitecore parallel publishing locks up Content Delivery server

Question

The original blog post on parallel publishing by Alex Shyba for SC 7.2 is a great resource and I think still applies to Sitecore 8.2.

I've dabbled with parallel publishing before on SC7.2 as well as just now on 8.2 and both times I've found that during a "large" publish operation (i.e. more than, say, 100 items) the busy CD server can, sometimes, completely lock up, rendering my website offline until I kill the app pool.

• Has anyone else experienced this?
• Why this might happen?

I suspect it's something to do with the shared core database where 2 instances (CM and CD) are fighting for locks on EventQueue rows but I've never really been able to get to the bottom of it.

No errors in any of the logs on CM or CD, no Windows Event Log entries, nothing I can see in SQL server (maybe I don't know what to look for, though).

The CM does not lock up - only the CD. The lock-up continues even after the publish has completed.

It's hard to test because it seems to only happen intermittently in a busy production environment which is hard to replicate.

The blog post above mentions enabling the Async EventProvider:

3. Enable the following /App_Config/Include file by renaming its extension to .config:

Sitecore.Publishing.EventProvider.Async.config.disabled

• Is this only required on the CM or on the CD as well?
• If not, could failing to enable the Async Event Provider on the CM somehow lock up the CD?

Answer 1

I think it's your big event queue. Try the following -

1. Make sure your CD indexes indexing strategy is set to manual. This will prevent double indexing of the web content. Of course there may be exceptions.

2. Another issue might be excessive processing of the event queue. Apply this patch - https://kb.sitecore.net/articles/930920 and see if it makes a difference. This will help a lot.

3. Keep the number of records in your eventqueue, publishqueue and history tables low by setting cleanup to every few hours. There is a patch somewhere which can allow you to trim your eventqueue down to minutes but I can't find it. You should have less than 10k items at worst in your eventqueue.

4. We had a problem where workflow publish was causing this issue for us (CDs locking up). If you are doing workflow publishing and have alllanguages, deep or related set to 1, then Sitecore by default does a republish of each item related to that instance of publish. We got a patch from patch from Sitecore that allows you to do smart publish of these items, essentially trimming the eventqueue massively.

I will try to add patches information when I'll be near a computer. Eventqueue is your problem so google event queue issues and you'll find a lot of people facing the same problem that you did.