On 8/7/23 14:52, Shamik Bandopadhyay wrote:
We are running a Solr Cloud infrastructure in our DEV environment with 4 Solr (nodes) and 3 zookeeper instances. Due to some terraform script issues, all three zookeeper instances were terminated. Once we restored them, the data (version-2 inside the data folder) was gone. Unfortunately, we didn't have any backup for the zookeeper data. Is there any way of restoring the Solr Cloud cluster using these newly created zookeeper instances without reindexing?
Whatever you do, don't fire up those old Solr nodes pointing to a new ZK ensemble. If you do that, Solr will proceed to delete all your index data.
First thing, make at least one additional copy of all the Solr data, in case something goes wrong during your recovery attempts.
Did you ever make a backup of the ZK structure or the ZK data directory? If you did, then you can probably restore the database with help from the ZK project, which should allow you to fire up the existing Solr machines without data loss and move forward.
Did you ever use the BACKUP action on the Collections API? If you did, then you can restore the collections to a new cluster with a new ZK database.
Failing both of those, you can probably create a new cluster and new collections with the same sharding setup, shut all the Solr nodes down, copy the index directories from the old setup over to the new cores, and start it all back up.
Thanks, Shawn
