Horton

Re: Improve LWLock tranche name visibility across backends

Sami Imseih <samimseih@gmail.com>

View thread View on postgr.es
From: Sami Imseih <samimseih@gmail.com>
To: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Cc: Nathan Bossart <nathandbossart@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2025-08-04T16:32:19Z
Lists: pgsql-hackers 
> > > I think we could add a local backend copy that stays up to date with the
> > > DSA. One idea would be to use an atomic counter to track the number of
> > > entries in the DSA and compare it with a local backend counter whenever the
> > > tranche name lookup occurs. If the atomic counter is higher (since we
> > > don't have deletions),
> > > we can update the local copy. Updating the local table should be a
> > > rare occurrence, but it would
> > > require an additional atomic fetch every time the name lookup occurs, in all the
> > > above code paths.
> > >
> > > Perhaps there's a better approach?
> >
> > I was thinking to switch to the DSA (and update local copy) when a name is
> > not found in the local copy. That way there is no need to maintain a counter and
> > the DSA overhead should be rare enough.
> >
> > Regards,
>
> That should work as well. good idea.

With a local hash table, I don't think it's necessary to introduce new
code for managing
a DSA based list of tranche names as is done in v3. We can go back to
storing the shared
trance names in dshash.

What do you think?

--
Sami

Commits

  1. test_dsa: Avoid leaking LWLock tranches.

  2. Teach DSM registry to ERROR if attaching to an uninitialized entry.

  3. Add a test harness for the LWLock tranche code.

  4. Revert recent change to RequestNamedLWLockTranche().

  5. Move dynamically-allocated LWLock tranche names to shared memory.

  6. Prepare DSM registry for upcoming changes to LWLock tranche names.

  7. Add GetNamedDSA() and GetNamedDSHash().