Technical Alert #7: Issue in shard synchronization.
On Thursday, February 3rd, we identified a critical issue in the shard replication mechanism that could lead to segmentation violations at runtime.
Do not roll out this version in cluster deployments. A fix is being prepared and tested. If you have already upgraded a cluster deployment to 3.8.5 and do not experience database server crashes, do not downgrade. Please install the fixed or any later version as soon as they become available.
An internal asynchronous call on shard followers accesses, in rare cases when shard synchronization with the leader is aborted prematurely, a voided register address leading to segmentation violation.
Review all database server log files of a cluster for spurious restarts, which were not commanded by database administrators. If and only if such restarts are found, please downgrade to 3.8.4 or any version available greater than 3.8.5.