Problem Definition

Current status Jira DC recovery from scratch:

1. Create a new instance of Jira node: node1
2. Deploy Jira to that node1 and connect to DB.
3. Start Jira
4. Since there is no index
   • you need to run Full reindex
   • or recover index from a snapshot.

This is a manual and time-consuming process. In case of AWS deployment with CloudFormation, when it's expected to redeploy the whole cluster for every upgrade problem becomes more important.

Suggested Solution

As a JIRA Datacenter Administrator I want to do an automated cold recovery from a snapshot. That includes the following steps:

• Periodically create index snapshot at Shared_home
• Add option (or JVM flag) for automatic index recovery during start-up
• If the option is set:
  1. search for index snapshot
  2. pause start-up
  3. restore index snapshot
  4. continue Jira start when the restore is done
• Log all the steps during the process and preferably expose status through REST API

Workaround

None

Related

• JRASERVER-66649 - to choose for index recovery vs full reindex
• JRASERVER-37896 - for standalone Jira

Note on fix

In 8.20.2 we changed the way we qualify snapshot as valid for being picked up on startup. Only snapshots created in less than 24 hours are considered valid.