Amanda, FreeBSD8, amtype, hairpulling, etc.

[Diego F. Arias R.]

On Sat, Mar 13, 2010 at 10:57 AM, Svein Skogen (Listmail Account) <
svein-listmail at stillbilde.net> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 12.03.2010 15:20, Svein Skogen wrote:
> > I'm having trouble getting Amanda (2.6.1p2 from ports) to play nicely
> > with my hardware.
> >
> > Devices are:
> > <HP Ultrium 3-SCSI Q25W>           at scbus0 target 1 lun 0 (sa0,pass0)
> > <HP 1x8 G2 AUTOLDR 2.80>           at scbus0 target 1 lun 1 (pass1,ch0)
> > connected via:
> > mpt0: <LSILogic SAS/SATA Adapter> port 0x9000-0x90ff mem
> > 0xfe4fc000-0xfe4fffff,0xfe4e0000-0xfe4effff irq 16 at device 0.0 on pci2
> > mpt0: [ITHREAD]
> > mpt0: MPI Version=1.5.20.0
> >
> > os and number in question is:
> > FreeBSD storage.stillbilde.net 8.0-STABLE FreeBSD 8.0-STABLE #0: Tue Mar
> >  9 07:01:59 UTC 2010
> > svein at storage.stillbilde.net:/usr/obj/usr/src/sys/GENERIC  amd64
> >
> > Tapes are regular LTO-3 (HP C7973A).
> >
> > amtapetype simply hangs (after writing 3-4 tps for 5 seconds, then
> > simple silence both on sa0 and console)
> >
> > Has anybody run into this problem with FreeBSD8+mpt+autoloader?
> >
> > //Svein
> >
>
> I'm finally starting to make sense of what I'm seeing (hence the
> crossposting)
>
> Seems I've stumbled onto some strange incompatibility between my SAS
> controller (LSI 3801E), Tape-library, and FreeBSD+Solaris.
>
> The behavior I'm seeing is consistent with this solaris bug:
>
> http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6894775
>
> but luckily (?) for me, my disks are on an MFI controller. I'm seeing
> the exact same behavior in both FreeBSD and OpenSolaris 133 and 134.
>
> Linux and Windows installs on the box (this box is currently being set
> up, so I'm rather liberated from reinstall-concerns) seems unaffected,
> atleast the HP Software doesn't fail the way
> tar/dump/dd/amanda/bacula/whatnot does at random intervals.
>
> The errors only occur when the device I'm reading/writing from is 100%
> laoded (reading or writing 56mb/sec + compression), which when fed from
> a raid capable of more than 6 times that is quite likely to happen
> during backups.
>
> The Solaris bug seems to be around MSI handling, but there are several
> reports "over there" about this error occuring even with MSI disabled.
>
> Right now I'm dumbstruck about this, and might install Linux "just to
> get backups up and running this year", because I'm too tired of this
> entire process, but I'd REALLY rather run FreeBSD or OpenSolaris. This
> is based on a personal preference and nothing else, but if anybody has
> some blinding insights on how to get this working, I'm open for
> suggestions.
>
> if there is a hardware/driver bug isnt possible to get another cheap sas
controller for your tape, maybe a second hand?
-- 
mmm, interesante.....