|From:||Nathan Bossart <nathandbossart(at)gmail(dot)com>|
|To:||Thomas Munro <thomas(dot)munro(at)gmail(dot)com>|
|Cc:||Kyotaro Horiguchi <horikyota(dot)ntt(at)gmail(dot)com>, pgsql-hackers(at)postgresql(dot)org|
|Subject:||Re: Suppressing useless wakeups in walreceiver|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
Here is an updated patch set with the following changes:
* The creation of the struct for non-shared WAL receiver state is moved to
a prerequisite 0001 patch. This should help ease review of 0002 a bit.
* I updated the nap time calculation to round up to the next millisecond,
as discussed upthread.
* I attempted to minimize the calls to GetCurrentTimestamp(). The WAL
receiver code already calls this function pretty liberally, so I don't know
if this is important, but perhaps it could make a difference for systems
that don't have something like the vDSO to avoid real system calls.
* I removed the 'tli' argument from functions that now have an argument for
the non-shared state struct. The 'tli' is stored within the struct, so the
extra argument is unnecessary.
One thing that still bugs me a little bit about 0002 is that the calls to
GetCurrentTimestamp() feel a bit scattered and more difficult to reason
about. AFAICT 0002 keeps 'now' relatively up-to-date, but it seems
possible that a future change could easily disrupt that. I don't have any
other ideas at the moment, though.
Amazon Web Services: https://aws.amazon.com
|Next Message||Tom Lane||2022-10-04 18:07:13||Re: Can we avoid chdir'ing in resolve_symlinks() ?|
|Previous Message||Jeff Davis||2022-10-04 17:39:31||Re: New strategies for freezing, advancing relfrozenxid early|