Column Store Deadlocks: Missing Information

People seem to care about deadlocks

That’s why I wrote sp_BlitzLock. Why not sp_BlitzDeadlock?

Well, I had this song stuck in my head. But enough about that.

While poking around with things recently, I created a pretty typical deadlock on a table, but this one had a clustered column store index on it.

Of course, that makes things different.

objectivity

So we go digging

The deadlocks I generated were on a single object. Just a couple inserts and deletes in two separate transactions. Nothing special.

This is the information that I pull out with sp_BlitzLock

Hey Mister DJ

But what’s missing from the column store deadlock information is really weird. Namely, everything.

Oooookay

One may be tempted to try to use the associated object id to get the object name, and one would be left wanting.

I wish you would.

And so, dear reader

I am currently unable to give you as much information about column store deadlocks as I’d like to.

Unless Microsoft responds to my Connect Feedback Item.

Brent says: this is such a great example of why niche features run into interoperability issues. They work just fine by themselves, but they’re not tested in combination with other things – even when those things might be mainstream, like deadlock graphs. I bet this bug gets fixed quickly though – with columnstore indexes gaining popularity, more folks will be running into this issue fast.

Previous Post
Building SQL ConstantCare®: 10% of you have Priority Boost on.
Next Post
How To Break SQL Server’s XML Data Collection

6 Comments. Leave new

  • “objectids” that are 17 digits long and start with 720575940 look like partition ids.
    This might give you something:

    select object_name(object_id) 
    from sys.dm_db_partition_stats
    where partition_id = 72057594059948032
    
    Reply
    • Erik Darling
      April 6, 2018 9:17 am

      Hi Michael!

      Yeah, I wanna say you can find them under allocation_units too, but that makes my life hard with more dynamic SQL. Information in the deadlock XML should be standardized and documented.

      Thanks!

      Reply
      • For sure, It should totally be standardized and documented. (I’ve already upvoted)
        As a blitz_lock author, you definitely need this fixed.
        I wrote the comment I guess for the onetime deadlock inspector who needs a workaround.

        Reply
  • I’ve up voted this because we’re using column store quite heavily and it would be nice to see it have feature parity with row store tables for things like this.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Fill out this field
Fill out this field
Please enter a valid email address.

Menu
{"cart_token":"","hash":"","cart_data":""}