Mboxlist DB Snapshot Rotation And Upgrade From 8.0.1 To 8.0.2

(Doc ID 2341032.1)

Last updated on DECEMBER 19, 2017

Applies to:

Oracle Communications Messaging Server - Version 8.0.2 and later
Information in this document applies to any platform.

Goal


If there is a problem with upgrading to MS 8.0.2 and we have to go back to 8.0.1, it is understood --per Doc ID 2323981.1-- that the following steps should be followed:

Then perform the downgrade. And after that, start stored and there should be no problem.

However, if there is a problem db, stored will try to use the snapshots.

The trouble is, all the snapshots are the same.

Note:  Prior to (sometime in early MS 7.x) the DB snapshots were used in a rotation. You could specify how many you want. It would keep track and always use the next one. So if you had 3 (the default) snapshots, you had 3 days of backups: the one from this morning at 02:00 (which would be basically "yesterday"), and the one from 02:00 the previous two mornings (ie, the day before yesterday and the day before that).

But that changed sometime in MS 7.x such that if you set the number of backups to 3 (still the default), the 02:00 job updates them all to be the same. So you now only have yesterday, regardless of what value you set for that config option (snapshotdirs).

 


The 02:00 imdbverify updates all the snapshots to be the same.
Then the every-10-minutes imdbverify updates the copy in the 001 directory.

After 02:00 the next morning, none of the snapshots will be usable if a down grade is necessary.

As a precaution, a backup could be taken of the 002 or 003 snapshot before the upgrade.

Or the following procedure can be used which is planned for the first one:

* stop-msg
* stored -r
* backup the mboxlist
* then upgrade

It is preferred to avoid having to do that on all of them. Once the first one is patched and no problems are found, then just patch as normal.
But the patching will be performed late at night. The 02:00 imdbverify will update the snapshots. So by the next day, when problems could potentially be found, the snapshots will already be gone. If a downgrade is needed and if stored -r and removing the logs was not sufficient (or someone forgot to do it), we would start with an empty mboxlist and have to reconstruct 25,000 accounts -- around 150TB -- which would take forever.

It is not expected to have to go backwards.
It is not expected that stored -r and removing the logs would be insufficient (or someone will forget to do it).

But if a problem is found and a downgrade is needed, it would cause panic. A massive reconstruct (or autorepair storm) would be bad.

Q1:  We would like confirmation and a recommendation to either:

Yes, do the same thing with every upgrade

or

Perhaps save one of the 02:00 snapshot copies. That would be about 20 hours out of date when it is captured--more like 2 days out of day if it had to be restored, but that would be infinitely better than starting with an empty mboxlist.

Q2:  What is the possibility for it to go back to the previous behavior where snapshotdirs = 3 means you have 2 days of backups, not 2 copies of the same thing from 02:00 this morning (what is the point of that)?

 

Solution

Sign In with your My Oracle Support account

Don't have a My Oracle Support account? Click to get started

My Oracle Support provides customers with access to over a
Million Knowledge Articles and hundreds of Community platforms