To summarise, recent posts could be categorised under the general heading of preventative measures:
- The relative merits and dangers of automatic v manual updates
- The use of a testing area (sandbox)
Both offer the system administrator practical means of (today) reducing the risk of a system failure as a result of updating it. Neither rely on extending the present functionality of eBox; they are examples of good practice.
Eventually, a system failure will occur. It then becomes a question of recovery measures, i.e. how the system is to be restored to operational condition. Two possible approaches to creating a fall-back position have been outlined:
- The use of imaging technology
- Rolling back the updates
Both offer the system administrator a means of reacting to unforeseen conditions.
Imaging can be used today; it does not require additional development of eBox. It is beyond the scope of eBox and might (quite properly) never be integrated with it. The example given in an earlier post illustrates its use on a machine removed from service during the upgrade. This might limit the appeal of imaging to some system administrators.
Rolling back the updates (as described earlier in the thread - see post #20) is the underdeveloped idea that was initially offered for discussion. If this is a workable idea, it will be operated manually from within the web-GUI (or automatically) and therefore will require input from the eBox team. It is not to be viewed as a replacement for good practice measures but is intended to to be used in conjunction with them.
In a situation where updating places eBox in an unresponsive condition, the eBox support team will be in exactly the same position as that faced should the condition occur today. Hopefully such a condition will be a rare and unusual event. When the machine remains responsive (following an update problem) the roll-back mechanism would confer the following major benefits:
- The system remains available to the end users
- The eBox developers and system administrator gain a time buffer to resolve the issue
- The eBox support team has a controlled mechanism (that they have designed and understand) which enables them to provide advice and rectify the matter more effectively.
Is this feasible?
