Translation of WriteConcernExceptions [DATAMONGO-1451]

[spring-projects-issues]

James Howe opened DATAMONGO-1451 and commented

Currently all WriteConcernException are translated to DataIntegrityViolationException, a NonTransientDataAccessException.

However, a write timeout also throws a WriteConcernException, for example:

com.mongodb.WriteConcernException: { "serverUsed" : "10.10.17.35:27017" , "ok" : 1 , "n" : 0 , "wtimeout" : true , "err" : "waiting for replication timed out" , "code" : 64}

Logically (IMO) this case should instead be a TransientDataAccessException, perhaps a subclass of TransientDataAccessResourceException or QueryTimeoutException. At least create a MongoWriteTimeoutException so I can special-case it in exception handlers more easily.

Unfortunately, the WriteConcernException doesn't appear to provide direct access to the wtimeout property, so I guess you'd have to use ex.getResponse().

My end-goal: I want to return 503 from my controller when it times out, and 500 for other non-transient errors

---

Affects: 1.8.4 (Gosling SR4), 1.9.1 (Hopper SR1)

Referenced from: pull request #370

1 votes, 2 watchers

[spring-projects-issues]

James Howe commented

This will probably require version 3.x of mongo-java-driver

[spring-projects-issues]

Mark Paluch commented

The difficulty is staying compatible with both driver versions, 2.x and 3.x

[spring-projects-issues]

James Howe commented

I'm also having similar issues with "Write failed with error code 100 and error message 'Not enough data-bearing nodes'".
Logically this would be the same as e.g. CassandraInsufficientReplicasAvailableException, another transient issue

[christophstrobl]

I don't think parsing the error message is the way to go here.
The MongoDB error_codes.yml holds a - RetriableError category for against which we could compare the actual MongoException.getCode() value.