0

Changing the configuration of a production server is dangerous.

A safety measure might involve taking a full system image snapshot prior to the change, changing the configuration and then, in the event of a severe problem, restoring from the snapshot.

However, during the interval the new configuration is being tested, various services might run that change the state in ways that should not be lost in the event or a restore. The number of different services running makes it difficult to journal each of them for restoration after the snapshot restore.

What is the best practice to handle this treacherous interval?

James Bowery
  • 158
  • 6

1 Answers1

0

You can't have your cake and eat it too. If you create a snapshot then you have to be willing to accept that whatever changes occur that you might want to keep will be lost in the event that you revert to the snapshot.

Your only other option is to take multiple snapshots to account for these changes and revert only to the snapshot that captured the specific state that you want.

This is the bargain. That the snapshot allows you to roll back unwanted changes while at the same time losing changes that might actually be desired.

joeqwerty
  • 108,377
  • 6
  • 80
  • 171
  • Of course you can't have your cake and eat it too. But what I'm asking for is the "best practice" which, I would presume, must include some set of procedures to, for example, disable certain services if not journal them, etc. so that snapshot restoration isn't _as_ destructive. – James Bowery Mar 02 '20 at 14:23