Re: SCSi issues after 10.0 update

[David Brownlee <abs%absd.org@localhost>]

On Mon, 10 Jun 2024 at 16:22, Riccardo Mottola
<riccardo.mottola%libero.it@localhost> wrote:
>
> Hi,
>
> David Brownlee wrote:
> > If you saw the initial issues with over NFS then bonnie++ over NFS
> > could be an excellent test
> >
> > (Many moons ago I used bonnie running on a handful of render clients
> > to show a proposed migration to Origin 2000 servers with Fore Systems
> > ATM drivers was going to be a disaster - repeatably triggering a panic
> > within a minute or so)
>
> I did connect a FreeBSD workstation running amd64 over wired ethernet to
> NFS serverd (same nfs that was in use when the issue happened). SO to
> use a reliable workstation and not other sparc clients, no doubts over
> doubts :)
> Stopped other clients... just not interrupt work. Serial console on the
> Netra T1 to see errors, logs, hangs.
>
> Run bonnie++ on the FreeBSD workstation on NFS mounted volume, with
> small size (128MB)- Starts to run but fails when removing directories:
>
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...Bonnie: drastic I/O error (rmdir):
> Directory not empty
> Cleaning up test directory after error.
>
> Actually, nothing was cleaned up.. I found the directory full of tmp
> files. Tried to delete it manually and it worked fine....
>
> Run bonnie++ on internal hard disks of the Netra, both disks work.
>
> I tried running boniee++ locally + remotely (even if it fails)... this
> should have tortured things a little bit, but no failure! No warnings in
> dmesg!
>
> Maybe it was really a bad day.. and a difficult to reproduce bug...
> however the fact that bonnie++ is not reliable over NFS isn't nice either.

That is unfortunate :/

When the issue triggered before which clients were busy - was it the
sparcs? I'm wondering if it may also be related to the specific use
patterns of those clients - also might be interesting to know if they
fail on rmdir the same way...

David