Rollback Strategy - Best Practices

[Former Member]

I would like to understand best practices followed for rollback in hybris deployments. hybris rollback seems to be not straight forward with no way to undo the 'update' process. With a rollback necessitates after few days of deployment, it can be a tough job. Any thoughts?

[Former Member]

Hopefully as part of your continuous integration and deployment processes, you would test/validate any type system changes which require an update to the system. These tests are typically done in lower environments (dev, staging, etc).

In a future version of hybris (see our Roadmap documentation on the wiki, this is still targeted for a Q3 2014 release) we'll have a Rolling Update feature, which will allow you to run your update on a single back-end node, validate everything is ok, and then have the other nodes in a cluster see the updated type system as part of a 'rolling' deployment. I would still employ deployment best practices and disaster recovery processes, have good backups ready in case there are issues, etc. Also make sure you have backups of your media file system, and restore to the last good point as necessary.

https://wiki.hybris.com/display/general/Roadmap

[christoph_meyer]

There are no provisions for a rollback.

Create a database dump after shutting down nodes for deployment and keep your old bin folder in case you need to roll back a deployment quickly.

If you perform schema updates that will prevent an old release from running (think of adding indices), you will have to deal with that manually in case of a later rollback.