Error 500 when deleting a db in a cluster without quorum

[jjrodrig]

Expected Behavior

When deleting a database in a cluster where the quorum is not met a 202 - Acepted should be returned if at least one of the active nodes responds ok.

It is related to issue #603

Current Behavior

When an existing DB is deleted in a cluster without quorum an error 500 is returned. The deletion is done and replicated once the cluster nodes come back.

Possible Solution

The problem seems to be here

couchdb/src/fabric/src/fabric_db_delete.erl

Line 73 in 1c39e0c

maybe_stop(W, Counters) ->

where accepted is returned only if quorum is met

Steps to Reproduce (for bugs)

1. Setup a 3 nodes cluster
2. Create DB - curl -X PUT "http://.../testdb"
3. Bring 2 of them down
4. Delete DB - curl -X DELETE "http://.../testdb"
5. A 500 error is returned
6. Get DB - curl "http://.../testdb"
7. 404 - Not found is returned

Context

The current situation is not consistent with the fix for issue #603

Your Environment

• Version used: Compiled from master
• Browser Name and version:
• Operating System and version (desktop or mobile): Centos 7
• Link to your project: