Re: Async client v4 mount results in unexpected number of extents on the server

[Benjamin Coddington <bcodding@redhat.com>]

On 12 Apr 2025, at 8:36, Holger Hoffstätte wrote:

> On 2025-04-11 15:29, Benjamin Coddington wrote:
>> On 10 Apr 2025, at 8:55, Holger Hoffstätte wrote:
>>> ...
>>> Does this behaviour seem familiar to anybody?
>>>
>>> I realize this is "not a bug" (all data is safe and sound etc.) but
>>> somehow it seems that various layers are not working together as one
>>> might expect. It's possible that my expectation is wrong. :)
>>
>> My first impression is that the writes are being processed out-of-order on
>> the server, so XFS is using a range of allocation sizes while growing the
>> file extents.
>
> Thanks for reading!
>
> Yes, that's true but not the cause of the re-ordering - the server just
> does what it's being told. The reordering has to come from the client in
> async mode.

There isn't any guarantee of write processing order for async writes, is
there?  I also don't think there's any practical impact.  So I'm wondering
what's the expectation of behavior and what problem you're trying to fix?

Ben