rcmart@rwc.urc.tue.nl (Mart Mennen) (06/07/91)
A few months ago we installed a SI59-Exabyte unit of the manufacturer System Industries on our DEcsystem 5400. The SI59-Exabyte is connected to the Q-bus via their QS1000 controller, which emulates a TK50. The Exabyte unit is driven by the standard DEC tmscp driversoftware. We are running Ultrix 4.0 and in addition to the SI59 we have 3 RA-disks, 1 TU81+ tape unit and 1 TK70 cartridge unit on the DECsystem 5400. These devices are configured in the Ultrix kernel as follows: adapter uba0 at nexus? adapter msi0 at nexus? adapter ibus0 at nexus? controller uda0 at uba0 controller uq0 at uda0 csr 0172150 vector uqintr disk ra0 at uq0 drive 0 disk ra1 at uq0 drive 1 disk ra2 at uq0 drive 2 controller klesiu0 at uba0 controller uq16 at klesiu0 csr 0174500 vector uqintr tape tms0 at uq16 drive 0 # TK70 device ln0 at ibus? vector lnintr controller klesiu1 at uba0 controller uq17 at klesiu1 csr 0160404 vector uqintr tape tms1 at uq17 drive 0 # TU81 controller klesiu2 at uba0 controller uq18 at klesiu2 csr 0160444 vector uqintr tape tms2 at uq18 drive 0 # Exabyte Until now we did not succeed to use the SI59-Exabyte without causing the system to panic once in a while! We use the "dump|dd"- and "dd|restore"-pipecommand and the tar command for writing and reading to/from the SI59-Exabyte. The problem does not show up immediately, but often occurs after successfully writing for 1-2 hours. The most occuring system panic was: cpu 0 panic: mscp_dealloc_rspid: sequence number mismatch Before this system panic a lot of mscp messages are reported in the errorlog, which all look like: mscp_message: dropping msg: mp c3d2d440, op a1, RSPID 4f60044 getpeername: Bad file number mscp - resynching controller uq18 mscp_dealloc_rsid: error rtp->rspid 50f0044, rp->rspid 4f60044 The remainder of the panics consisted of: cpu 0 panic: trap cpu 0 panic: unaligned access cpu 0 panic: m_free has bad m_cltype In the meantime the firmware of the QS1000 controller was upgraded, but didn't solve any of the above problems! At the moment we don't dare to use this Exabyte unit anymore.... So, we are interested to know if there are any users who have succeeded to get the SI59 to work on a DECsystem 5400 or similar DEC system. Perhaps someone has a clou why it doesn't work on our system. Secondly, are there users of DECsystem 5400's, which successfully use Exabyte units of other vendors? If this is the case, could you give me some information about these vendors and their Exabyte product(s)? Thanks in advance for your time and help. -- Mart Mennen, Eindhoven University of Technology Computing Centre 1.99, +31 40 472164 P.O.Box 513, 5600 MB Eindhoven, The Netherlands E-Mail: rcmart@urc.tue.nl
GEOMAGIC@kcgl1.eng.ohio-state.edu (Daniel OConnell) (06/12/91)
>A few months ago we installed a SI59-Exabyte unit of the manufacturer >System Industries on our DEcsystem 5400. >Secondly, are there users of DECsystem 5400's, which successfully use >Exabyte units of other vendors? >If this is the case, could you give me some information about these >vendors and their Exabyte product(s)? >Thanks in advance for your time and help. We have a CMD 223 controller in our 5400 working with an Exebyte. However, working means only that dump and restore, and tar work. dd operations can fail occasionaly, but dd also fails on our 9-track and TK-70 since upgrading (downgrading ?) to Sultrix 4.1. This class of programs has been verified at other Ohio State Univ. sites and is apparently an operating system device driver bug or something along those lines. I have hope (faint) that this is partially resolved in Sultrix 4.2, but reading the release notes does not encourage much hope. Meanwhile, a $240,000 seismic reflection processing software package is disabled until this TMSCP tape problem is fixed. It works with SCSI tape drives that the software vendor uses. Of course, all our three 5400 tape controllers make tapes (including SCSI Exebyte) look like DEC TMSCP tapes so we would have optimal compatibility with DEC TMSCP tape bugs... sigh. Cheers, Dan O'Connell geomagic@geo1s.mps.ohio-state.edu