Being able to run two separate couchdb instances on the same machine is considered a feature. Not being able to tell them apart is considered a bug (in wetware).
:) B. On 31 August 2011 21:38, Markus Jelsma <[email protected]> wrote: > Hi, > > Incredible. I somehow managed to start two instances running on the same port > with a different bind address?!? > > There's only one log file and the test DB isn't listed in > /var/lib/couchdb/0.11.0. Yet i can query the test DB using localhost but not > using the FQDN. > > I've an import running on the FQDN now so when it's finished i'll stop/kill > all instances. Using lsof i now see the test DB as test.couch (deleted). > > Is it considered a feature to be able to end up in this mess - starting two > instances but getting different data depending on bind address? I now remember > that the restart of the first running couch on localhost failed and actually > expected it to continue gracefully. It didn't stop the first running couch but > did start the second on the new bind address. > > Thanks > >> It sounds as though when you access localhost you’re reaching a _different_ >> CouchDB instance. Are you sure the machine doesn’t already have a CouchDB >> running, perhaps bound only to the loopback interface? For instance, GNOME >> desktop, at least on Ubuntu, uses a local CouchDB to manage the user’s >> address book/calendar data. Maybe you’re hitting that one. >> >> Try looking at <http://localhost:5984/_utils/> and see if your missing >> documents are showing up there. >> >> —Jens >
