awpsys@ultb.UUCP (Andrew W. Potter) (10/17/89)
> > We are running a VAX-6300 with HSC70 to control the disks. We are now >considering connecting one of the MIPS machines (5800 series) to the HSC. >Now, there are two questions: >1. Is it possible at all? (ofcourse each will use different disks). Our > salesman says it'll work. >2. A more general question: Is it possible to limit inside the HSC access to > some disks for some hosts; i.e.: disk 0 to 7 will be access only by host A > and disks 8 to 10 only by host B? I couldn't find anything in the > HSC manual, so is it impossible? > Thanks in advance, > __yehavi: I have been running ultrix and VMS over the same CI for almost a year now. It works like a champ. DEC does not support the configuration because of the millions of ways you can shoot your self in the foot. Basicly you must teach both VMS and Ultrix to STAY AWAY from each others disks. Since neither VMS nor Ultrix understand or respect each others file system volume protections, you as a system manager MUST take steps to prevent users from scribbling on the disks of the other OS. Specifically: VMS will allow a user to "$ INITIALIZE" a mounted and running ultrix file system and Ultrix will freely allow someone to "newfs" a mounted and running VMS system disk. Here is how you prevent such disasters: Designate which spindles will be ultrix and which will be VMS. On VMS in the system startup file on do a $ SET DEVICE/NOAVAILABLE to every device that will be accessed by Ultrix. On ultrix remove the device specifications for ALL VMS disks from your /usr/sys/HOSTNAME config file. Also delete all the special files for the VMS disks from /dev. If possible AVOID the use of /genvmunix and the canned DEC ultrix installation procedures. these files will gleefully find and select any and all devices on the CI. It is best to build your CI based unix kernel before booting into the CI. If you MUST use the DEC supplied procedures for Ultrix installation then I HIGHLY recommend shutting down your entire VMS cluster and WRITE PROTECTING all VMS disks. Note that Ultrix does not support HSC allocation class. Ultrix will see HSC000$DUA1 and HSC001$DUA1 as two separate disks. Select the path that preferred and leave the other drive port select button de-selected. (You can do a manual HSC failover by having an alternate ultrix kernel configured using the other path and reboot on alternate HSC if there is a failure). The HSC can STILL have an allocation class set for use by VMS. If it does, then you must remember to disable the Ultrix disks from VMS specifying the allocation class) Having taken all the above precautions, you can then run and update systems pretty much normally without having to worry about them. The benefits are as follows: VMS and Ultrix can share Tape drives: (Just remember to REMOVE your write rings when you want to MAKE SURE that your tape is not going to be written on.) We use a procedure that switches access between ultrix and VMS by alternately closing the Unix permissions on the tape drive special file and $ ALLOCATing the device from VMS using our operator console account. If you have DEC field service and run VAXSIM, then you can benefit from VAXSIM reporting and analyzing errors on the Ultrix disks. If you are getting a high error count on an ultrix disk, you can quickly recover by: a) Shutting down Ultrix b) SET DEVICE/AVAILABLE ultrix_disk: c) ALLOCATE ultrix_disk: d) MOUNT/FOR/NOWRITE ultrix_disk: e) MOUNT/FOR/NOWRITE spare_disk !(of same geometry) f) BACKUP/PHYSICAL ultrix_disk: spare_disk: g) DISMOUNT/NOUNLOAD ultrix-disk: (and spare_disk:) h) de-select A and B ports from Both disks i) change unit select plugs between the Ultrix disk and the spare disk: j) SET VOLUME/NOAVAILABLE ultrix_disk: Since BACKUP/PHYSICAL can copy a disk in about 15-30 minutes you can get the system back up in relatively short time compared to the traditional Unix way of dump and restore. This method requires that the spare disk be on the same HSC as the ultrix disk. If you are really clever you can actually make the Unix systems read-only accessable to VMS. Instead of the SET DEVICE/NOAVAILABLE, you do a MOUNT/SYSTEM/FOREIGN/NOWRITE/PROT=W ultrix_disk: /NOWRITE is MOST IMPORTANT. /PROT=W prevents casual programmers writing a program to read it You can then write programs to read the unix file system. (with the cavaet that it may not be fully up to date due to unix buffer caching. - Andy -- Andrew W. Potter Bitnet: awpsys@ritvax.BITNET Systems Programmer Internet: awpsys%ritvax.bitnet@cunyvm.cuny.edu Information Systems and Computing Rochester Institute of Technology, Rochester NY, 14623 (716) 475-6994