Community resources
Community resources
Community resources
- Community
- Q&A
- Confluence
- Articles
- So you want to upgrade from 9.2.x to 10.2.14
So you want to upgrade from 9.2.x to 10.2.14
This is not strictly a question, but more so an aggregation of problems i ran into when upgrading our confluence cluster. Some have been solved, some not (yet). If you have any ideas/tips for the latter ones, they are greatly appreciated; but mostly this is for some other poor sod that runs into the same problems and could use a helping hand (or word for that matter).
So, the usual backups, pg_dumps etc. aside i have an ansible playbook that takes care of downloading the confluence zip, adjusts a few variables and - a new addition - takes care of the oh-so-lovely UPM truststore certificates. Don't even think about starting your instance without them if you ever looked in the direction of the atlassian marketplace. Further instruction on this topic can be found here: https://confluence.atlassian.com/upm/configuring-upm-app-signature-check-1489470544.html
Having learned from our staging environment i was confident that our (numerous) addons wouldn't make a problem. "How naive" one might think - and one would be right.
After starting confluence.service for the first time i was met with a wall of ERROR messages in the log files, mostly related to the Osgi Plugin being unable to resolve... well basically anything.
But don't fear, there is always the handy "--disable-all-addons" flag for start-confluence.sh (start the script as the confluence user - else you'll end up with files/folders confluence probably won't be able to access).
[edit]
Almost forgot! If you happen to run confluence in a cluster: that does not get along with 'disable-all-addons' (because why would it?). But don't fret, there is the 'confluence.cluster' property in confluence.cfg.xml (to be found in the home a.k.a. data folder). Set it to 'false' and you're good to go - just don't forget to set it back to true after you're done.
[/edit]
After successfully starting confluence it was a matter of upgrading all our addons to the newest version and providing about half of them with their license keys (which had gone missing mysteriously).
We shut down confluence with stop-confluence.sh and tried the systemd.service once again. This time successfully: confluence was up and returned the so much sought after "RUNNING".
First server up and running i updated the other two and they promptly popped up in the clustering list. What else could go wrong you might ask? Hold on to something sturdy, we're going for a ride:
- on the License Details Page the server ID shows up as $action.serverID - easily fixed manipulating the confluence database: https://confluence.atlassian.com/confkb/server-id-shows-as-action-serverid-action-sid-or-n-v-181536632.html
- Scroll Translations - activated for some spaces, not for others (and yes, i'm only talking about spaces that used it before the upgrade). This as well is a rather simple fix: just go the space in question and activate Scroll Translations - if you know which spaces should use it... Similar in origin to the $action.serverID - values from BANDANA table somehow don't make it to the plugin_setting table.
- The config of our OnlyOffice Document Server for our OnlyOffice Connector
- The settings for Team Admin that define what groups to auto-create in a new space
And now for the really fun ones (a.k.a.: the problems i opened a support ticket for):
- those User macros you so lovingly handcrafted in velocity? They're gone. And i don't mean "they don't render properly", nonono. The "installed User Macros" list is wiped clean. Nothing. Nada.
- and last but in no ways least: you've heard about those "Spaces", yes? Wouldn't it be awesome if we could make more of those? Tough luck, unknown server error says "no". The logs say "java.lang.NullPointerException: date must not be null"
I've found several similar articles on the topic - the one that got closest is here: https://support.atlassian.com/confluence/kb/creating-new-page-or-space-loads-forever/ but unfortunately none of the solutions fit.
As i've already said: some of these problems have already been solved, the rest hopefully will be soon with the help of atlassian's support team. I will aim to [update | comment on] this once the solutions present themselves in case somebody else needs them.
[edit, 22.06.2026]
concerning the creation of new spaces: the plugin_setting table struck again (that little bundle of data sure causes a lot of trouble). In the BANDANA table there is an entry for the server installation date (bandanakey = confluence.server.installation.date) which apparently did not get transferred.
First solution attempt: restart confluence with '-Datlassian.forceSchemaUpdate=true' (e.g. via setenv.sh). Did not work in our case, but maybe it will do the trick for somebody else.
Oh and a fun little bonus: our users can't create pages from templates either. Same 'date must not be null' in the logs so i assume that problem has the same origin.
[edit, 23.06.2026]
partial success! Thanks to atlassian support i now know the magic words in postgresql to make confluence believe it was installed six years ago:
INSERT INTO plugin_setting (namespace, setting_key, setting_value) VALUES ('_GLOBAL', 'confluence.server.installation.date', '#java.lang.Long 1577836800000');
The numerical value is the first of January 2020 in epoch time which works for our needs - adjust if necessary.
Still, my goal is to make the forced schema update work. So far there are some entries missing from plugin_setting that i know of because something obvious broke. I suspect there are more things missing that will bite me in the behind a couple weeks/months.
*added the OnlyOffice/TeamAdmin items to the list*
Was this helpful?
Thanks!
Ulrich Lazar
3 comments