Hello everyone,
as Confluence recently get zero downtime upgrades for feature releases (https://jira.atlassian.com/browse/CONFSERVER-52686) I was thinking again if I really like ZDU independent from the marketing phrases. It sound absolutly good to reduce downtime / impact for end-users specifically on large instances but on the other hand it gives you new challenges.
And the main challenge in my opinion is then consistent backups and in the worst case the restore when things go wrong.
The internal upgrade tasks itself become more stable in the past but still there is the chance that it might fail.
With Confluence 7.13 and some garbage data (Gliffy) in bandana table the upgrade task for XStream was failing and the instance stopped immediatly. So these cases still happen these days.
What we need for a consistent backup (example Confluence):
- Database which is shared (e.g. size of 160GB)
- Home directory (e.g. size of 10GB)
- Shared home directory (e.g. size 800GB)
Depeding on the size backups are challenging to keep everything consitent also with incremental backups.
Putting the Confluence into read-only mode would mitigate and help to backup the data nearly consistent when everything is running but it's also an impact to end-users as they can't edit / comment anymore.
So what's you opinion on ZDU and backups?
