Thread Deadlock : Solved

This blog is the solution to the earlier stated problem : Solve the Deadlock

.

 

 : Read the error message carefully. Its more helpful than we tend to think.

.

 

The error tells us that the problem happened during the update of the ‘person’ object.

.

So, let us start looking at the lines that update the ‘person’ object. A bit of digging gives us … line #28 and #37. That is …

person_queryset.update(**kwargs)

.And

Person.objects.update(session=Subquery(mapping_queryset))

.

Since these are the only lines that update ‘person’ object. So, we can be (reasonably) sure that these lines are the cause of the problem

.

 

If I knew nothing about the python programming language or the framework used. Then, I would think that line #28 updates a sub-set of ‘person’ objects.

.

person_queryset.update(**kwargs)

.

And line #37, updates another subset of ‘person’ objects.

.

Person.objects.update(session=Subquery(mapping_queryset))

.

It might be possible for each DB update to be going via a different worker thread. The threads might be locking on different tables/ rows … causing the deadlock.

.

Turns out – this in fact was the problem. !

 

.

But, the method annotation did cause me some confusion. Methods marked transactional usually use a single DB transaction for DB updates.

.

However, each library/ framework might implement transactional behavior differently. (I was discounting this fact )

 

.

  : Each library and framework implements a functionality differently.

.

The method updated a fixed number of ‘persons’. And did NOT need to update different sets.

 

Knowing this, line #37 was re-written from:

Person.objects.update(session=Subquery(mapping_queryset))

.to :

person_queryset.update(session=Subquery(mapping_queryset))

.

Voila ! problem solved.

.

.

Shoutout to the guy who figured this out. You rock !

.

.