Using Postgres advisory locks

[bymyslf]

Is there anything that prevents the use of the advisory locks instead of the lock table?

IMO, it would be better to use the native functions. What do you think?

https://www.postgresql.org/docs/9.4/static/explicit-locking.html

[ahydrax]

Hi @bymyslf , there are several things preventing from using advisory locks:

• pg_advisory_lock accepts only integer values whereas hangfire needs a string for the key
• advisory locks are binded to current session whereas hangfire lock owner can release the lock via different connection/session

I had a long journey investigating the most "natural" way to implement distributed locks in pg and finally I came to this but it's applicable for pg versions 9.6 and above.

[bymyslf]

Hi @ahydrax, the first point is quite embarrassing for me. I was fully confident that the job ID was used as the resource and didn't check twice. Confused with another project that uses distributed locks.

However, even if the job ID was the resource, your second point invalidates the "native" locks.

Thank you for your thoughts.