Technical Alert #5: Possible data loss for collections created with v3.4.6
On Monday 3rd June, we identified a critical issue in our latest 3.4.6 release of ArangoDB. Customers that run ArangoDB Clusters are affected. Single Instances and Active Failover setups are not affected.
The agency, which closely manages the cluster communication might – under certain circumstances – mark newly created collections for deletion.
This is a serious issue that needs urgent action!
See the issue resolution description below to solve potential issues with newly created collections and upgrade to the 3.4.6-1 patch version afterwards
A bug in the handling of time-to-live (TTL) entries in the cluster’s agency, can lead to complete data loss for all newly (after upgrade to version 3.4.6) created collections. The failure case can be triggered as a consequence of leadership change among the agents.
A fix is available in the form of a repair script to allow an upgrade without losing collections and the new ArangoDB v3.4.6-1 release with a permanent fix for the underlying problem. We provide detailed instructions for manual cluster deployments, Starter deployments as well as for Kubernetes in a separate article:
Please avoid adding new collections as long as your cluster runs on the unpatched version 3.4.6.