1

I have an instance on which Vespa is installed. This system is acting as admin as well as content node and cluster node. We are also using the embedded zookeeper in Vespa for our use case.

Requirement: How can we take a backup (of documents and zookeeper node) in such a way that if there is an issue on the instance we can use the backup on another instance to clone the documents. (Only one instance at a time for Vespa is available. So we won't be able to create another group with a replication factor).

Yash Kasat
  • 203
  • 1
  • 5

1 Answers1

1

As Vespa distributes documents using buckets over all nodes in a group (and the replicas generally hits different groups), the best way is to dump the full corpus from all nodes using https://docs.vespa.ai/documentation/content/visiting.html. You can conceivably stop a content node and copy the index files, but this error prone.

As Vespa auto redistributes buckets of documents, the best approach for a failing content node is to add a new one and retire the failing using vespa-set-node-state

zookeeper is used by the config server and cluster controller. To back up its zk data, you should stop the config server, then copy from the latest snapshot in $VESPA_HOME/var/zookeeper/version-2

Also see https://docs.vespa.ai/documentation/cloudconfig/configuration-server.html

Kristian Aune
  • 876
  • 5
  • 5