AI Query

[ron]

We are about to introduce AI for the first time.

We have a production server (P) and a backup server (B) that is used for nothing at all except to be ready to take-over if needed.

Scripts are been written to switch to the next AI extent on P for each of five DBs every (say) one hour ... then rcp them to B server ... and apply them to the five DBs.

Our wish is to keep the servers on P “up” all the time. B will have servers “down” all the time ... unless that system has to take over.

At 18:00 each day we intend stopping applying AIs to the B databases - and then backing-up the whole of system B.

After the backup, will Progress allow further AIs to be applied? Presumably the AI extents from P will continue getting higher extent numbers - where on B (presumably) Progress will expect AI extents to start at 1 again.

How do we co-ordinate the databases on P and B so that the AI files can be applied to B after a backup?

 

[toby.Harman]

I think you have an assumption here......

Progress doesn't restart the ai counter unless you re-implement ai. By that I mean re-enable ai - I.e. rfutil aimage end, full backup and then rfutil aimage begin. That is the only way I know of to get the counters back to zero.

So at 18:00 the AI files start queuing up to be applied to the "B" system.

When doing the backup remember the -no recover option if using probkup. This will prevent the backup software from going through crash recovery and changing your "B" databases.

Quick question - You have 5 "A" databases and 5 "B" databases. Does anything write to more than one database in a transaction? If so - are you using 2 phase commit? How will you manage the AI files that are not precisely in sync (though they will be largely!)

Would it be possible to reduce the number of databases and use storage areas or is there a need to manage the databases seperately (e.g. -B for each one etc)? It would relieve this problem of synchronising databases.

Alternately we can all lobby Progress for -B tunable by db area as well as by db

 

[ron]

You're saying, in effect, that we have no problem - so long as we set-up the backup process correctly. We'll proceed on that basis.

Our system is a package - so we have no control over the way transactions are handled. But although the system deals with 5 DBs at the same time ... to the best of our knowledge no transaction would update more than 1 DB.

Thanks a lot for your help!

Ron Hunt.

 

[toby.Harman]

Well - I won't say no problems but.... :awink:

The only one I can immediately think of is that you were quite correct in identifying the AI counter as being an issue. This number does have a top limit (and I can't remember what it is but something like 32767 or 65535).

When your AI extent reaches that number it will either roll over to 0, 1, -32767, -65535, or your database will lie down and die!

When you get close to that point in time you'll need to do an AI end, a backup, an AI begin and restore that backup to your recovery server.