Handler errors should be formatted as JSON.

[pires]

• CHANGELOG.md updated
• Rebased/mergable
• Tests pass
• Sign CLA

Fixes #3170

[otoolep]

Thanks @pires

Now that I think about this, this may be considered a breaking change, we need to be careful about this.

@pires -- can you show us an example of the before-and-after for this change?

[pires]

@otoolep same error code but the only difference is that the error is JSON encoded as per issue request.

Is there any reasoning behind this not being the case before?

[pires]

Previous behavior:

$ curl -v -X POST 'http://192.168.99.100:8086/write?db=notexist'
*   Trying 192.168.99.100...
* Connected to 192.168.99.100 (192.168.99.100) port 8086 (#0)
> POST /write?db=notexist HTTP/1.1
> Host: 192.168.99.100:8086
> User-Agent: curl/7.43.0
> Accept: */*
>
< HTTP/1.1 404 Not Found
< Request-Id: a57fcaeb-95a7-11e5-8003-000000000000
< X-Influxdb-Version: 0.9.3
< Date: Sat, 28 Nov 2015 08:11:37 GMT
< Content-Length: 31
< Content-Type: text/plain; charset=utf-8
<
database not found: "notexist"
* Connection #0 to host 192.168.99.100 left intact

PR behavior:

$ curl -v -X POST 'http://localhost:8086/write?db=notexist'
*   Trying ::1...
* Connected to localhost (::1) port 8086 (#0)
> POST /write?db=notexist HTTP/1.1
> Host: localhost:8086
> User-Agent: curl/7.43.0
> Accept: */*
>
< HTTP/1.1 404 Not Found
< Content-Type: application/json
< Request-Id: 156d97ff-95a8-11e5-8001-000000000000
< X-Influxdb-Version: 0.9
< Date: Sat, 28 Nov 2015 08:14:45 GMT
< Content-Length: 45
<
{"error":"database not found: \"notexist\""}
* Connection #0 to host localhost left intact

[otoolep]

Agreed @pires -- it's the same error, just wrapped in JSON.

@pauldix -- I don't think we ever meant to send back non-JSON formatted errors, so this makes sense to me. But this is technically a breaking change. Need your call.

[pires]

@otoolep thank you. I think the question about breaking changes begs for more reviewing of the currently open issues, so to prevent anyone else spending time with stuff that in the end may get rejected. Agree?

[otoolep]

Yeah, makes sense @pires -- I'm on the fence about this one, since it's looks like a bug that it was never JSON-format.

[benbjohnson]

I'm personally ok with this breaking change. It seems like it should have been JSON to begin with.

[pires]

Any updates on this?

[pauldix]

Yep, should have been json to begin with. Thanks @pires!