Restore GS160 on a 4100

[Wim Van den Wyngaert]

For a restore test, I should restore a complete 7.3 GS160 system on an old 4100.

Anyone did that already ? Problems (except disk layout) ?

Wim

[Ian Miller.]

VMS V7.3 runs on AS4100 just fine.

[Wim Van den Wyngaert]

No problems restoring the system disk of a GS on the AS4100. Reconfig needed ?
(never had to do this before, only 4100 to 4100)

Wim

[Volker Halle]

Wim,

OpenVMS is OpenVMS is OpenVMS !

You can take any OpenVMS system disk from any Alpha system to any other Alpha system and just boot it...

Assuming that:

- the system is supported by the version of OpenVMS on that disk
- properly written startup command files (use STARTUP_P1="MIN" for the first boot)

Volker.

[Bill Hall]

Wim,

The only other often overlooked hardware issue is memory. You may need to due a fair amount of sygen parameter tuning if your production server has much much more memory than your 4100. If you have an Oracle instance that wants 4GB of memory for SGA and you are testing on a server with only 2GB of memory, you will have some additional software config changes to make.

The sysgen and application changes are usually harder to identify and quantify in the planning stage than the more obvious hardware differences that can be "hidden" with logical names.

[Tom O'Toole]

The point that Volker makes is a huge one.

People spend countless hours and dollars trying to do proper disaster recovery. What is so difficult on other systems which seem to lock into the hardware and specific config. they are running on, is easy on VMS. The marketeers writing white papers on about ROI, TCO, will of course ignore this, as it doesn't fit their agenda, but how much are people spending purchasing and configuring 'enterprise backup solutions' and server virtualization on non-vms boxes, and how well do they work?

[Wim Van den Wyngaert]

I would go further. How many companies have a tested procedure of what to do in case of total loss of all disks ?

How to continue an application based upon the backup of the previous day ?

Wim

[Uwe Zessin · Accepted solution]

You might need different licenses for VMS.

There might be differences in device names:
- disks -> use lots of logical names
- network
-- DECnet lines + circuits
--- MOP download circuits
-- TCP/IP interfaces
-- LAT links
-- Pathworks interface
-- LANCP
--- MOP downloads

[Uwe Zessin]

> How many companies have a tested procedure of what to do in case of total loss of all disks ?

Not many I beleive. I've worked in several companies and usually inherited an environment.

In one of them I accidently discovered that the incremental backup had been saving the second disk twice rather than saving all 3 disks for several months!

In my last position as a full-time VMS system manager I spent quite some work on a 'system handbook' which contained all kinds of information - from hardware/software configuration, our own system procedures, recovery procedures and troubleshooting information.

On one monday I had to decide to scrap a whole cluster! One of the members had a bad CPU which caused strange errors: for example, I had a file named ABC.LO - the uppercase "G" was replaced by a . As it was unclear where else the CPU had corrupted data, I stopped the bad member and used the intact member to reload all disks including the system disk (we had a clever boss who budgeted(sp?) for a spare disk!) The system was several hundred kilometers away, I did not have to travel and could start the restore immediately. Thanks!.

We were lucky, because the last full backup was from friday, no work done on the weekend and we discovered the problem moday morning, pheww.

> How to continue an application based upon the backup of the previous day ?

Good question. Today's backup tools are quite capable and if you strictly follow the established rules you have a high change to successfull do a desaster recovery. The problem is how to restart from the RP and redo the manual work.

In some cases, companies run multiple backups during the day or use technlogy like log shipping to keep the RPO as short as possible.

[Jan van den Ende]

Wim,

you know the stories on our cluster.

Our test/plat/sometimes development system _IS_ a 4100.

From our point of view the most important uses are:
- testing any major updates/upgrades. OS as well as Packages and applics.
- running What-if scenario's
- (connected to your question) testing our Restore.

Granted, the latter is not rigidly scheduled, but each time a testing or what-if has been done, that 4100 gets a restore of our fill backups.

Yes, it only has 1 CPU
Yes, it only has 1 Gb
Yes, the shadow sets are single member.

Most importantly, ALL configurables are contained in only 3 files, AND, virtually this system is a full member of the cluster. (ie, it has its own root, with the correct SYSGEN params. It can, and has, joined the cluster as a full member. The only thing we do to set it apart from the cluster is setting VMSCLUSTER = 0.

If we want to/need to "refresh" our test system, we typically restore the backup and boot. That _IS_ all!

And it HAS helped us greatly when we needed to rerun a specific batch of a specific applic over the situation of 6 weeks before.
It only took 1 morning to "time-warp" (and another to get back in sync).

That was one event in which we were able to turn a REAL embarrasment for management into just some overtime cost.

To get back to your question:

Anyone did that already? To a 4100, yes. From a GS? no, but from a cluster of ESes.

But, to smoothly do it, you should prepare for it.

Success!!

Proost.

Have one on me.

jpe