Date: Thu, 30 Jan 2014 17:44:41 -0700 From: "Justin T. Gibbs" <gibbs@scsiguy.com> To: Ben Laurie <ben@links.org> Cc: freebsd-scsi@freebsd.org Subject: Re: Dropped interrupts Message-ID: <80532F4C-9206-4268-936E-66F352087052@scsiguy.com> In-Reply-To: <CAG5KPzzrenHbYSBY5zQ1GNKg%2BT2nY0aAKp3uGSw=Rxn1=KYnow@mail.gmail.com> References: <CAG5KPzxUAnORPSyJDrSdBBNE7MBi-dD=M6=E1rMG%2Bc8rn4deUQ@mail.gmail.com> <AE41EB93-26BD-4BD0-9913-7CD0A5B6E1A4@scsiguy.com> <CAG5KPzwR5=WNKc5hck8F7CtCtk3mivwYFRFeCJT_zWdnetW=3w@mail.gmail.com> <84D23688-DDC6-421E-9D21-3DA646229038@scsiguy.com> <CAG5KPzxqxmzfnAiUyVyApMs5XqKpH0F0sbTxy1h=4cL=rEROpQ@mail.gmail.com> <CAG5KPzyPXLEOAFKTAW%2BJaai=XbT3mi5yGHs1LnUbMcD-LfQ1YQ@mail.gmail.com> <CAG5KPzzVpvH=miguqKnO0pA2jmWm-_7s7W-Nbh%2BVeHKv7x7ikw@mail.gmail.com> <8680A21D-78D7-4B65-A502-17F0C3B70291@scsiguy.com> <CAG5KPzwvBoBvh210Vn5GG6eBVNXKWCkuh6J_FLGao0L6aPAmcw@mail.gmail.com> <888F991A-96C7-48C0-AAC6-0D432BD78A46@scsiguy.com> <CAG5KPzzrenHbYSBY5zQ1GNKg%2BT2nY0aAKp3uGSw=Rxn1=KYnow@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Jan 30, 2014, at 3:21 PM, Ben Laurie <ben@links.org> wrote: > On 26 January 2014 22:46, Justin T. Gibbs <gibbs@scsiguy.com> wrote: >> On Jan 26, 2014, at 1:52 AM, Ben Laurie <ben@links.org> wrote: >>=20 >>> On 25 January 2014 17:15, Justin T. Gibbs <gibbs@scsiguy.com> wrote: >>>> On Jan 24, 2014, at 10:44 PM, Ben Laurie <ben@links.org> wrote: >>>>=20 >>>>> Aha, finally got the error again... >>>>=20 >>>> I don't know enough about your backup or Bacula to know if this is = amount that should be written, but we attempted a write in variable = block mode of 64512 bytes. >>>=20 >>> I don;t think there's anything wrong with that. Not at the machine >>> right now, but I think that's actually the default max block. No = idea >>> why. >>>=20 >>>> The command, data transfer list, and controller state are all = consistent with this. The command was successfully transferred to the = tape drive, but it never transitioned to data phase to allow us to begin = the data transfer. (In parallel SCSI, the target controls all state = transitions). >>>>=20 >>>> Since there are no parity errors or other indications of a = transport error, my hunch is that this is a tape drive issue. Are you = running the latest available firmware for it? >>>=20 >>> No idea, its a very old LTO-2 drive. I'll see if I can find an = update. >>>=20 >>>> How many write cycles do you have on your media? >>>=20 >>> I have seen this error on brand new media. I'd guess the most cycles >>> any tape has had is around 10. >>>=20 >>>> When was the last time you cleaned the drive? >>>=20 >>> LTO drives are self-cleaning. So ... never. >>=20 >> This is a popular misconception. LTO drives include brushes deployed = during load or unload to remove large particle contamination off the = head. This increases the time interval between traditional cleanings, = but doesn't make them unnecessary. Certain LTO drives will tell you = (e.g. light a LED) when they require cleaning, but I don't know if this = is one of them. There have also been firmware bugs on some drives that = prevent the cleaning indicator from working as expected. >>=20 >> If the drive has over 500 hours on it, I would suggest a cleaning = pass. >=20 > Are you suggesting that a need for cleaning could cause this problem? I can=92t say why the drive isn=92t responding. But if it hasn=92t been = cleaned since it was manufactured, you are probably getting less data = per tape than optimal, slower transfer speeds than optimal, or both. I don=92t know if the LTO-2 model listed on this page is the same as = what you have, but if it is, there were several firmware releases after = 3AYC: http://www-01.ibm.com/support/docview.wss?uid=3Dssg1S4000055 =97 Justin=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?80532F4C-9206-4268-936E-66F352087052>