home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.sys.sun.admin:10757 comp.periphs.scsi:6708
- Newsgroups: comp.sys.sun.admin,comp.periphs.scsi
- Path: sparky!uunet!munnari.oz.au!uniwa!bilby.cs.uwa.oz.au!dunnart!janet
- From: janet@cs.uwa.oz.au (Janet Jackson)
- Subject: Help! Worrying behaviour of Exabyte 8500C
- Message-ID: <janet.728032198@dunnart>
- Summary: retries on dumps; strange behaviour of LEDs; write errors
- Sender: usenet@bilby.cs.uwa.edu.au
- Nntp-Posting-Host: dunnart
- Organization: Dept. Computer Science, University of Western Australia.
- Date: Tue, 26 Jan 1993 07:09:58 GMT
- Lines: 134
-
- (Please excuse the crosspost. I'm not sure whether it's a SunOS software
- or Exabyte hardware problem.)
-
- Environment: Third-party Exabyte 8500C (5Gb 8mm drive with hardware
- compression interface) connected to Sun 4/470 server running SunOS 4.1.1.
- I mostly use it for dumps.
-
- The Exabyte has been acting up lately. There are three main symptoms,
- which may or may not be related. They all seemed to start happening
- at about the same time, though.
-
- The first symptom is strange. The second symptom is REALLY strange.
-
- FIRST SYMPTOM:
-
- The first symptom I noticed is error messages like:
-
- Jan 26 14:17:44 wambenger vmunix: st2: warning, the tape may be wearing out or
- Jan 26 14:17:44 wambenger vmunix: the head may need cleaning.
- Jan 26 14:17:44 wambenger vmunix: st2: write retries= 685 (4.1%), file= 3, block
-
- on just about every dump. The dumps are written OK. I get read retries on
- restores, too.
-
- Each night my automatic backups cat a short ascii label to the front of the
- tape, then dump 29 filesystems. The cat doesn't give any error messages,
- but each dump after that gives two: one each for the first and last block of
- the file. The number of retries increases from near-zero to 32000-odd, at
- about the 25th file; after that it's large negative numbers and ridiculous
- percentages, like 700% -- presumably this is overflow. This seems to imply
- that the errors are not real, but that some error counter is being incremented
- incorrectly.
-
- Dumps I've done manually only seem to give one error message -- perhaps the
- double messages are an artifact of the automatic backup software.
-
- Once the error has started happening to dumps on a given tape, it continues
- to do so until the tape is ejected. After that, if you use the same tape
- again, it may work OK for a while, but once the error occurs once it
- will happen again on subsequent dumps, again, until you eject the tape.
- It's as if there's some sort of status in the drive that gets reset when you
- eject.
-
- Cleaning the heads does not help.
- Replacing the tape does not help either, although in one of my tests,
- a new tape did not exhibit the problem the first time I dumped on it.
- This has not generally been the case, though.
-
- Before you say "FAQ": I installed patch 100134-03, which is supposed to
- stop spurious messages of this sort, but it hasn't helped.
-
- As far as I can tell it is only dump that produces these messages.
- cat, dd, cpio and tar all work normally, on both new and "second-hand" tapes.
- (Maybe because of different blocksizes?)
-
- These messages are the least annoying of the symptoms, since the dumps
- do work. However:
-
-
- SECOND SYMPTOM:
-
- This is the most worrying of all.
-
- Normally when the Exabyte is writing to the tape, the amber LED flashes a
- lot, indicating SCSI activity (according to the manual), and the green
- LED occasionally flashes slowly, indicating normal tape motion.
-
- However, I have lately noticed that during the occasional dump (no more than
- one in 30) the green LED flashes slowly most of the time, and the amber
- LED only occasionally. When this is happening the dump runs about 30 times
- as slow as usual. It's as if it's looking for a good place to write on
- the tape! However, just because one dump on a given tape does it, doesn't mean
- that another one will.
-
- I have no idea under what conditions this occurs. I certainly can't _make_
- it happen! It's happened on various tapes (new and old; different brands);
- it's happened just after I cleaned the heads as well as when I hadn't cleaned
- them for ages; other activity on the SCSI bus doesn't seem to make any
- difference.
-
- I've only ever seen this happen during dump, but then I hardly ever use
- tar, cpio, etc on this tape drive.
-
-
- THIRD SYMPTOM:
-
- We have had an unusual amount of write errors (hard errors, causing dumps
- to fail) lately. By "unusual", I don't mean it fails constantly -- I mean
- three in the last month. One of them seems to show that it is not the tape
- that's at fault: on 23/01 a write error occurred 17 feet into the first dump
- on a tape. I tried the SAME TAPE again for last night's dumps, and no error
- occurred.
-
-
- WHAT'S CHANGED?
-
- I first noticed these problems in late December. What have I changed
- shortly before that time?
-
- - I've installed patch 100570-03, an ethernet controller patch (surely
- irrelevant!)
-
- - I used some new tapes of a different brand: Verbatim DL112M, rather
- than our usual Sony QG-112M.
-
-
- NOTES
-
- The Sun 4/470 doesn't have an 8500 driver. It thinks the drive is an 8200,
- but that's never seemed to make any difference to its operation.
- Should I install a proper driver? If so, where can I get one?
-
- I've only ever used data-quality tapes in it, never video tapes.
- I clean the heads regularly with an Exatape cleaning cartridge from Exabyte.
-
-
- SO WHAT'S WRONG?
-
- Has anyone come across this before (especially the second symptom, the LEDs)?
-
- Do you think the heads are damaged? The compression interface, perhaps?
- Some other internal component? Could changes in humidity or in the power
- supply affect it? Could having a heavy Sun Desktop Storage Module on top
- of it be getting to it, even though it has worked perfectly in this
- condition for almost a year?
-
- If you have any thoughts about this, please email them to me.
- I will summarise if there's interest.
-
- Janet Jackson
- <janet@cs.uwa.edu.au>
- Systems Administrator
- Department of Computer Science
- The University of Western Australia
-