jim@baroque.Stanford.EDU (James Helman) (09/19/90)
Two out of the last thirty tapes we've gotten from SGI have been bad. I just spent most of the day trying to track down a replacement for the IRIX 3.3.1 Maintenance Tape which arrived this morning. I was enthusiastically installing it (because I thought it contained a bug fix I need) when the installation bombs in midstream. Our 4D/220's 150MB drive encountered errors part way through the tape. Repeated attempts failed. Ditto on a 4D/80's 60MB drive. I try everything. It turns out a Sun-3's ancient 60MB drive (which regularly backs up a gigabyte of disk and has never been cleaned or maintained) can read the tape without errors, so I finally dd'ed the distribution onto a new tape and finished the aborted installation. Still, it's kinda inconvenient when an installation fails and leaves your machine with a partial brain transplant. Now, 2 out of 30 is a 7% failure rate. With 6 tapes in average distribution, this would imply a 36% chance of at least one tape in the distribution being bad! So either I have really rotten luck, two bad tape drives, or SGI has a big problem someplace. Anyone else looking forward to CD distributions? Jim Helman Department of Applied Physics Durand 012 Stanford University FAX: (415) 725-3377 (jim@KAOS.stanford.edu) Work: (415) 723-9127
fsfacca@AVELON.LERC.NASA.GOV (Tony Facca) (09/19/90)
> >Two out of the last thirty tapes we've gotten from SGI have been bad. > ** stuff deleted ** > >Now, 2 out of 30 is a 7% failure rate. With 6 tapes in average >distribution, this would imply a 36% chance of at least one tape in the >distribution being bad! So either I have really rotten luck, two bad >tape drives, or SGI has a big problem someplace. > Pick any two?? Though I haven't kept close track, I'd say your numbers are right on target with what I've been experiencing. In the 3.3 release (about 7 or 8 tapes) the DEV tape was bad. This can be really frustrating since it leaves the system pretty confused. After a few shell escapes, a couple of 'versions remove' commands, and a new tape everything comes together, but.. >Anyone else looking forward to CD distributions? Sign me up. -- ----------------------------------------------------------------------------- Tony Facca | fsfacca@avelon.lerc.nasa.gov | phone: 216-433-8318 ----------------------------------------------------------------------------- You are at Witt's end. Passages lead off in *all* directions.
karron@MCIRPS2.MED.NYU.EDU (09/20/90)
I too have had bad tapes from sgi. I usually end up getting a new tape from them. Still, don't they try to re-read the tapes, or otherwise verify tapes after they make a copy ? dan. +-----------------------------------------------------------------------------+ | karron@nyu.edu Dan Karron | | . . . . . . . . . . . . . . New York University Medical Center | | 560 First Avenue \ \ Pager <1> (212) 397 9330 | | New York, New York 10016 \**\ <2> 10896 <3> <your-number-here> | | (212) 340 5210 \**\__________________________________________ | +-----------------------------------------------------------------------------+
bobg@rains.wpd.sgi.com (Bob Green) (09/22/90)
In article <9009192232.AA23148@mcirps2.med.nyu.edu>, karron@MCIRPS2.MED.NYU.EDU writes: |> I too have had bad tapes from sgi. I usually end up getting a new tape from |> them. Still, don't they try to re-read the tapes, or otherwise verify tapes |> after they make a copy ? |> |> dan. |> +-----------------------------------------------------------------------------+ |> | karron@nyu.edu Dan Karron | |> | . . . . . . . . . . . . . . New York University Medical Center | |> | 560 First Avenue \ \ Pager <1> (212) 397 9330 | |> | New York, New York 10016 \**\ <2> 10896 <3> <your-number-here> | |> | (212) 340 5210 \**\__________________________________________ | |> +-----------------------------------------------------------------------------+ Each tape that is produced by SGI or its vendors is rewound and read for data consistency. On a sampled basis tapes are moved from one drive to another to verify tape drive head alignment. Our current research shows one cause above all others in contributing to tape read errors. We have found that foreign material on the tape drive heads will increase the probability of read errors. In most cases this results in recoverable errors which just slow the read process. In a few cases it results in unrecoverable tape read errors. The solution in all cases is to clean the tape drive heads. We have included a request in our installation instructions for users to clean their drives BEFORE attempting installation. In a few cases (0.01% of tapes produced) we find that even with a clean tape drive that there is an actual tape error. We are researching these to find what in our processes could be causing this. A third possibility is that the tape drives were not calibrated correctly when manufactured. Although we have not seen evidence of this, we are keeping our ears open. Bob Green Software QA Mgr Silicon Graphics, Inc (bobg@sgi) (415) 962-3438