The NFS server that stores ES indices got disconnected from my server while my provider was doing some maintenance. This corrupted two indices:
.kibana_task_manager_7.12.1_001 and a custom one called
applications. The former seemed to be completely unrecoverable, but some searching suggested Kibana would just re-create it if I deleted it, and that has indeed moved my status from red to yellow and the system is usable again.
The other one, though, has two copies, and
_cluster/allocation/explain says of one of them "a copy of this shard is already allocated to this node..." - and it does indeed list two indexes, one unallocated.
I should be able to just delete the unallocated one: the allocated one seems to be working (presumably was recovered; but I could re-create it if necessary from backup). But
DELETE /applications is ambiguous, as I have two. How can I delete just the unassigned one? If I did do
DELETE /applications with cURL, would it delete both? Or just the assigned one? For obvious reasons, I don't want to just try it! If I can avoid deleting the assigned one, that would be helpful.
The info from
explain goes on to give
node [[admissions], node[dlSVRrRuTFyqeSCzepCtjA], [P], s[STARTED], a[id=kGzjo8mzTlqg0GOPNBCRHg]]" and
"node_id" : "dlSVRrRuTFyqeSCzepCtjA" so presumably that would allow me to identify the problem index, but what request can I use to do the delete?
(The server is running a single shard, everything is on the one server, no replication).