FAQ - Why do records disappear when a node is brought into a cluster?
Why do records suddenly get removed from a cluster when a node that was formerly part of this or another cluster is added? The newly-added node has had its data disks completely wiped, so it shouldn’t have anything that affects data.
Although the records themselves live on the devices specified in the configuration file, there is some important information also kept in the system metadata (SMD) directory. This includes potential truncate times (before which records should be removed) for sets and namespaces that have had truncate or truncate-namespace performed on them.
Use caution when returning a node to the cluster. This phenomenon can result in unexpected data loss. Deleting all files in the SMD directory before returning the node and letting them be repopulated from other nodes in the cluster is the safer option.
SMD TRUNCATE TRUNCATE-NAMESPACE REUSE RECYCLE RETURN BRING BACK NEW NODE WIPE ZEROIZE DISK