[OpenAFS] moving a RW volume which several RO copies : when the vos release becomes a "complete release" ? addenda

[Michael Meffie]

On Thu, 27 Feb 2025 11:24:11 +0100
Giovanni Bracco <giovanni.bracco@enea.it> wrote:

> Hello!
> Recently, when upgrading the OS of a file server, we have moved only the
> RW volumes to an already upgraded fileserver, leaving the RO copies on
> several sites where they were.
> 
> When the upgrade was completed we have moved back the RW on the original
> fileserver.
> 
> RW Volumes where NOT modified in the meanwhile but when issuing a "vos
> release" it becomes nevertheless a "complete release" and this is
> annoying as the sites connections have an rtt of about 5 ms and rx
> performance is limited to about 10 MB/s.
> 
> So the question is:
> 
> why the  vos release becomes a full release if the RW volume is just
> moved between fileservers, not modified anyway?
> 
> Sorry I have not specified the OS and the OpenAFS versions
> 
> the upggrade has been from CentOS7.9/openafs 1.8.11
> to Almalinux9.5/openafs 1.8.13
 
Hello Giovanni,

Sorry about the painful release. Do you have the text of the vos output?

I suspect the full release was triggered because the read-only clone was
missing since the read-write volume was moved.  Ever since OpenAFS 1.0,
vos will fall back to a full release when the read-only clone is missing
when the release is started.  In general, vos release tends to be
conservative and falls back to a full release to cover corner cases
to make sure the release is complete.

Mike --

-- 
Michael Meffie <mmeffie@sinenomine.net>