Re: [PATCH v6 00/12] core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objects

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: May 24, 2022, 12:31 UTC
Message-ID: <nycvar.QRO.7.76.6.2205241425570.352@tvgsbejvaqbjf.bet>
In-Reply-To: <fb1f2c8c-b749-85d9-b472-587f51dea8a8@gmail.com>

Hi Neeraj,

On Thu, 19 May 2022, Neeraj Singh wrote:

Show 24 quoted lines

> On 5/19/2022 2:47 PM, Junio C Hamano wrote:
> > Junio C Hamano <gitster@pobox.com> writes:
> >
> > > nksingh85@gmail.com writes:
> > >
> > > > From: Neeraj Singh <neerajsi@microsoft.com>
> > > >
> > > > GGG closed this series erroneously, so I'm trying out
> > > > git-send-email. Apologies for any mistakes.
> > > >
> > > > This series is also available at
> > > > https://github.com/neerajsi-msft/git/git.git ns/batched-fsync-v6.
> > > >
> > > > V6 changes:
> >
> > We haven't heard anything on this topic from anybody for this round.
> > I am planning to merge it to 'next' soonish.
> >
> > Please speak up if anybody has concerns.
> >
> > Thanks.
> >
>
> No updates on my end. I'll keep my eyes out for any reports of regression.

I asked a colleague to have a go with these patches and the only concern I heard back was that with ext4's new `fast_commit` feature, the `fsync` seems not to actually flush all metadata. They indicated that they'd be happy with merely documenting this issue, and also pointed out that the `fast_commit` feature seems still not to be considered ready for production workloads.

So: 👍 from my side.

Ciao, Dscho