SUMMARY: Strange Exabyte 8505XL problems

From: Nick Murray (nmurray@csd.abdn.ac.uk)
Date: Wed May 19 1999 - 02:51:22 CDT


It turns out the main problem was self-inflicted! When I moved the tape
drive it switched from /dev/rmt/1* to /dev/rmt/0*. My home-grown dump
script already had a /dev/rmt/0* entry so I just uncommented it, not
realizing it was /dev/rmt/0ln, hence the low read/write rates!

 It seems that the fresh tape I was using got labelled as low density
and hence I got the low data rates even using the correct parameters
on the command line during my testing yesterday.

Thanks to:

Tim Carlson
Bob Rahe

Original question:

> Hi,
> I'm having a strange problem with a Sun 14GB 8mm tape drive
> (Exabyte 8505XL) connected to an Ultra Enterprise 250 system. The
> tape drive used to be connected to a SS1000, but it was moved to the
> UE250 yesterday morning.
>
> The problem is that on the UE250 the device only manages about
> 250KB/sec reads and writes (this is irrespective of blocking factors,
> compression, etc), whereas on the SS1000 it managed about 750KB/sec.
> I noticed this as the nightly backup was taking a very long time.
>
> The SS1000 was running Solaris 2.6 and the device was connected to a
> fast/narrow SCSI interface. The UE250 is running Solaris 7 and the
> Exabyte is the only device on the external Ultra/Wide SCSI bus -
> connected via a 68-50 pin SCSI cable. The Ultra/Wide SCSI bus has been
> used before for Fast/Wide devices and the cable has successfully been
> used before, but not both together.
>
> The "strange" thing is there are no errors logged, no retries from
> ufsdump and scsiinfo just reports:
>
> glm1: st4,0 tgt 4 lun 0:
> Synchronous(5.000MB/sec) Clean Narrow
> Removable Tape: EXABYTE EXB-8505SMBANSH2 07T0 [S]
>
> The only thing amiss is that the throughput is down by two-thirds!
> >From a perfmeter & iostat output it looks like the device is active
> for 1 second and then busy for the next 3 giving a very distinctive
> perfmeter trace.
>
> So, has anyone seen anything similar? It is likely to be the the tape
> device or even the cable? - although with duff cables/devices I've
> always seen errors logged!
>
> Thanks in advance,
>
> +-------------------------------------------------------------------------+
> |Nick Murray email <nmurray@csd.abdn.ac.uk> |
> |Senior Computer Officer phone (+44 1224) 272307 |
> |Dept. of Computing Science Fax (+44 1224) 273422 |
> |University of Aberdeen |
> |King's College |
> |Aberdeen AB24 3UE |
> |Scotland UK |
> +-------------------------------------------------------------------------+



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:20 CDT