[comp.os.vms] re LAVC and rd53

UHAC010@vaxa.rhbnc.ac.UK (07/03/88)

You _can_ run an LAVC off an RD53 boot disk. I have just brought up a
VS2000 on a Q3 microvax with twin RD53's. However, to do this I had to use 
a dump of the system disk from my other (already running) LAVC which does have 
an RD54, since as you rightly say the installation procedure for VMS V4.6
refuses to continue in the prescence of an RD53 system disk. My 
understanding of this restriction is as follows.

 VMS 4.6 is a monolithic group of files. MicroVMS is essentially the same 
set of files (minus some important header files, examples and other things 
that _real_ VMS programmers value) but in five lumps, namely the BASE 
system and four `options' called PROG, SYSP, USER and UTIL. DEC rather 
grandly told us all how a microVMS system was `configurable' ie you could 
have a system with no programmer commands, no system programmer support, no 
users (heaven really does exist) and no utilities. Dunno quite what you use 
such a beast for, presumably embedded applications, but then I thought 
VAX-FLAN was the solution to all those problems. In addition to these 
options, DECnet (which is part of the monolithic big VMS) is just another 
layered product on MicroVMS.

 The reason for all this partitioning is that with the advent of the 
microVAX DEC found itself selling systems with disks that were too small 
for real VMS. The problem was exacerbated by the way that VMS installs 
itself. Loosely speaking, a version of VMS is loaded from the tape and then 
it builds the new system. At critical points in the installation process 
two versions of VMS are effectively resident on the disk at the same time, 
so you need LOADSA-blocks. Micro-VMS takes the central chunk of VMS that 
actually needs to participate in this process and calls it the base system. 
Only after the rebuilding is completed do you actually load the rest of 
the OS, so all the blocks that would have been filled up by things like the 
security utilities, queuing processes etc. are avaliable for the build, and 
hence it can be done on a microVAX.

 Now DEC never bothered to build a microVMS system that could support 
clusters for reasons that are unknown to me but probably have something to 
do with the fact that space on an RD53 really is a bit tight even if you 
can get it all to work. I have not yet finished purging back the new system but 
I only had about 10K blocks left and that was with page and swap file sizes 
set too small for comfort.

 The good news is that at V5 microVMS dissapears because big VMS has 
adopted the BASE+OPTIONS strategy and can therefore be tailored for small 
disks. However RD52 disks are no longer supported atall because they are 
just too small. RD53s are fine. I've never used a microVAX I with an RD52 
but from what I know of the performance of the machine and its disk I 
should imagine anybody out there with such a combination will be glad of a 
reason to junk it.

 The way to get your RD53 LAVC going is to find someone with a live LAVC 
and then do a BACKUP/IMAGE/VERIFY/IGNORE=INTERLOCK of the system disk. Boot 
standalone backup on your boot node and restore from the tape using
BACKUP/IMAGE/VERIFY mua0:*/save dua0:/INIT. This takes about an hour and a 
quarter. Then reboot the machine off dua0. You had better do this with your 
Ethernet cables disconnected since the machine will think it is the same 
DECnet node as the one you got the tape from, and that might cause some 
confusion... Immediately run NETCONFIG.COM to put you network to rights and 
then run ncp to enable service on your Ethernet adapter (as per p10 of the 
VMS LAVC manual, AA-JP20C-TE). Run BOOT_CONFIG.COM and then reconnect the 
Ethernet. You can then run SATELLITE_CONFIG.COM to add in your shiny new 
VS2000 (or whatever).

 When I did this I took the precaution of using SATELLITE_CONFIG.COM to 
remove all the satellites from the system disk of running system before I 
took the backup. I think you could easily run into space problems on your 
RD53 if you omit this step.

 Words of warning.
 1) Your satellites should have local disks for page and swap because there
probably isn't enough space on the RD53 for them. 

 2) You'd better have another disk to put you user files on because there 
won't be much space on the RD53. The satellite local disk would be fine.

 3) When your boot node is on the net with no satellites awake you will get 
regular messages from DECnet complaining about aborted service requests. 
This is normal, don't worry. As soon as the cluster reaches quorum they 
will go away.

 4) DEC do NOT approve of this procedure. I am an edu user with a known 
penchant for hacking my systems around and so they decided not to notice 
what I am doing but don't be surprised if they are unhelpful when you 
complain to them that this doesn't work


 If you want more details please contact me directly. Noone in the states 
ever seems to be able to contact me. Anyone on the UK side know what my 
mail address looks like to US readers?

------------------------------------------------------------------------------

Adrian Johnstone,

Mail:	Royal Holloway & Bedford New College, Egham, Surrey, TW20 0EX, England
Phone:	0784 39025
Fax:	0784 37520
Telex:	935504 RHC
Email:
 Janet:	A.JOHNSTONE@uk.ac.rhbnc.vaxa 
 Arpa:	A.JOHNSTONE%vaxa.rhbnc.ac.uk@cs.ucl.ac.uk  (I think)
 Bitnet/NetNorth/Earn:	A.JOHNSTONE@vaxa.rhbnc.ac.uk  (I think)

------------------------------------------------------------------------------