Feedback-ID: id4a34324:Fastmail
Date: Fri, 9 Jun 2023 21:35:38 -0700
From: Andres Freund <andres@anarazel.de>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Christoph Berg <myon@debian.org>, Thomas Munro <thomas.munro@gmail.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: could not extend file "base/5/3501" with FileFallocate():
 Interrupted system call
Message-ID: <20230610043538.2z46xyoaohj4fjnf@awork3.anarazel.de>
References: <20230425001623.im5uquriqlk4ayy6@awork3.anarazel.de>
 <20230606195300.wackhn5qc7nic5l5@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230606195300.wackhn5qc7nic5l5@alvherre.pgsql>
Archived-At: 
 <https://www.postgresql.org/message-id/20230610043538.2z46xyoaohj4fjnf%40awork3.anarazel.de>
Precedence: bulk

Hi,

On 2023-06-06 21:53:00 +0200, Alvaro Herrera wrote:
> On 2023-Apr-24, Andres Freund wrote:
> 
> > A prototype of that approach is attached. I pushed the retry handling into the
> > pg_* routines where applicable.  I guess we could add pg_* routines for
> > FileFallocate(), FilePrewarm() etc as well, but I didn't do that here.
> > 
> > Christoph, could you verify this fixes your issue?
> 
> So, is anyone making progress on this?  I don't see anything in the
> thread.

Thanks for bringing it up again, I had lost track. I now added an open items
entry.

My gut feeling is that we should go with something quite minimal at this
stage.


> On adding the missing pg_* wrappers: I think if we don't (and we leave
> the retry loops at the File* layer), then the risk is that some external
> code would add calls to the underlying File* routines trusting them to
> do the retrying, which would then become broken when we move the retry
> loops to the pg_* wrappers when we add them.  That doesn't seem terribly
> serious to me.

I'm not too worried about that either.


Unless somebody strongly advocates a different path, I plan to push something
along the lines of the prototype I had posted. After reading over it a bunch
more times, some of the code is a bit finnicky.


I wish we had some hack that made syscalls EINTR at a random intervals, just
to make it realistic to test these paths...

Greetings,

Andres Freund