fail-over options for Confluence Wiki

[Wikified]

What options are available for adding fail-over capabilities to Confluence? I'm not worried about scalability, more about have a recently synced copy of the Wiki available in case our production Wiki isn't available.

Has anyone tried using MySQL repliction?

Is there a VM or 'snapshot' solution available?

[Andrew Frayling · Accepted solution]

Hi,

I've not used MySQL replication, but I have implemented Confluence with a cold-standby for failover with Oracle dataguard and that worked well enough as manual failover with a short period of downtime (1 hour) to carry out the failover was acceptable. For that implementation we stored all of the attachments in the database so didn't need to worry about sync'ing the attachments to the failover server or back again.

We used a developer license (http://confluence.atlassian.com/display/DOC/Getting+a+License+for+a+Staging+Environment) for the cold-standby and the 2 Confluence nodes were fronted by a Netscaler load balancer with traffic to the standby node disabled.

I've also implemented failover with both Confluence nodes pointed at the same database, but making sure that only one node was active at any one time and didn't get any problems with data corruption or cluster panics. That implementation also used Oracle dataguard to sync the database to a backup, but we could failover Confluence nodes without also having to failover the database.

I'd avoid VM snapshots for backups or failover as that's not what they're designed for, but if you're talking about running Confluence in a virtualised environment so you can migrate VM's if there are problems with the host then that should be fine (http://confluence.atlassian.com/display/DOC/Running+Confluence+in+a+Virtualised+Environment)

Hope that's of some help?

Andrew.

[Eric Seibert]

Everyone is interested in 2 nodes to one Database with an inactive one. 

We are working on a similar configuration.
1 Netscaller/LB running with both nodes, the ssl cert resides on the LB
2 nginx/httpd has the same configuration on both nodes.
3. rsync process to keep home and install directories in sync and/or a shared mount for shared files
4. Node 2 has a stopped instance of Confluence/Jira/...  and a CI/CD tool monitors the instance running on both nodes if it finds no confluence running it starts the second one (thats the tricky part )

realistically it would be ideal to have data center and be done with it but my institution is bedded with another intranet technology and doesn't understand the value  and opportunities of Atlassian product offerings and integrations

I'm interested to see other solutions there seems to be lots of conjecture about this methodology but no consistent execution. I'm sure its due to the nature of the Atlassian server vs data center model

[Wikified]

All good advice. Since my fail-over wiki only needs to be current within 1 day I'm thinking of doing a mysqldump from the production wiki to a network drive and then importing that to the faili-over wiki once a day.