couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Newson <>
Subject Re: view_indexer fails / cryptic error messages
Date Fri, 22 Mar 2013 12:58:56 GMT
Docclass::map_doctype #version=19">>,

doesn't look quite right to me. Could you friendpaste/gist your design document?

FYI: If you try uploading your design document to Cloudant you'll get
an error message if it's malformed or the map/reduce code does not
compile. I think we're porting that work as part of the merge, or
shortly after.


On 22 March 2013 08:38, Dave Cottlehuber <> wrote:
> On 22 March 2013 09:20, Dave Cottlehuber <> wrote:
>> On 22 March 2013 06:16, Gerald Richter - ECOS <> wrote:
>>> Hi,
>>> I am using couchdb for over two years now and I really love it, but I hate it’s
error messages. If there is one thing I would wish that should be in the next release, than
it would be: human readable error messages. I know that writing good error messages is nothing
that programmer likes to do. We like more to implement great new features, but for the daily
work having a good diagnostic message is of so much value and saves so much time…
>> Good call, maybe you want to open a JIRA for this and we can discuss
>> what needs to change?
>>> So in addition to the many hour I spend in the past, trying to decrypt an error
message of couchdb, here is one that really gives me no clue what it is about and I am giving
up. Neither starting couchdb with debug output, nor looking at the output of my viewserver,
shows anything irregular. Basically a view index is rebuild and it crashes somewhere in the
middle of the process with the error message below.  I hope somebody on the list, could shed
some light what’s going on here.
>> If you can run the view build with log level = debug  and send a link
>> to the results we might see something more interesting.
>> Don't forget what version of erlang, and couch you are running.
>> I'm pretty sure one of these fields in that log excerpt will tell you
>> the last doc processed but I don't see it right now myself.
>> If your view is small, you might consider sprinkling in some log
>> messages
>> bearing in mind there's a *significant* performance impact for doing
>> this.
>> If your view is failing at the same place each time, you'll see this
>> when you delete the view, re-run the build, query _info (see below)
>> and then re-run the build and re-query _info again.
> I forgot the most important bit (aaugh), in that view _info output is
> the update_seq  which should match the db's info at
> $COUCH:5984/dbname/  if it completed all docs. If it didn't you should
> be able to GET $COUCH:5984/dbname/_changes?since=<update_seq> and one
> of the docs around there  should be the culprit. You can use
> limit=5&descending=true as well to get a list of the preceding docs,
> and just limit=5 to get subsequent doc ids. I'm not 100% sure that
> this all will work, maybe one of the other devs can confirm, as there
> are a few circumstances[1] where the update_seq might not match
> between view & db, but it should be reasonably close.
> [1]
> A+
> Dave

View raw message