Concurrency on indexing

[MatteoPiovanelli-Laser]

This error looks similar to what we worked on for concurrency:

ERROR GenericADOException thrown from IContentHandler by Orchard.Indexing.Handlers.CreateIndexingTaskHandler
NHibernate.Exceptions.GenericADOException: could not execute query
[ select indexingta0_.Id as Id5888_, indexingta0_.Action as Action5888_, indexingta0_.CreatedUtc as CreatedUtc5888_, indexingta0_.ContentItemRecord_id as ContentI4_5888_ from Orchard_Indexing_IndexingTaskRecord indexingta0_ where indexingta0_.ContentItemRecord_id=@p0 ]
  Name:p1 - Value:Orchard.ContentManagement.Records.ContentItemRecord
[SQL: select indexingta0_.Id as Id5888_, indexingta0_.Action as Action5888_, indexingta0_.CreatedUtc as CreatedUtc5888_, indexingta0_.ContentItemRecord_id as ContentI4_5888_ from Orchard_Indexing_IndexingTaskRecord indexingta0_ where indexingta0_.ContentItemRecord_id=@p0] ---> System.Data.SqlClient.SqlException: La transazione (ID di processo 170) è stata interrotta a causa di un deadlock delle risorse blocco con un altro processo. Ripetere la transazione.

[MatteoPiovanelli-Laser]

Note that this often happens when indexing is enabled on a tenant, and we try to upload several images at the same time from the MediaLibrary UI

[MatteoPiovanelli-Laser]

This created a funny situation, being that there was an indexing task for a ContentItemRecord that did not exist, and that caused every BatchIndex execution to fail and log the same error over and over.

[MatteoPiovanelli-Laser]

I gave quick look at how the IndexingTaskRecord is used. It's always through an IRepository, and we never call an Update. This is most likely one of those cases we can fix easily with the NOLOCK stuff we did in dev. I'll figure out a way to reliably cause this issue and get back.