Robert, > On 12 Dec 2015, at 18:02, Robert Samuel Newson <rnew...@apache.org> wrote: > > Hi, > > Cloudant has recently deployed a CouchDB 2.0-based build into an internal > environment (it gets real usage but is not yet on our public clusters) and > we’ve found it to be pretty solid. I note a few things that we saw (they > might not all apply to 2.0, we extend and modify the code a little);
this is outstanding news, thank you! \o/ :) > 1. now that we’re clustered, it’s important to remember to set [couchdb] uuid > to the same value for each node a cluster (and for different clusters to have > different values). At Cloudant all clusters have names, so we just use that > (if you’re familiar with our offering, values like "meritage", "moonshine", > "jenever"). > > 2. View-based filtered replication seems not to work, probably a simple > oversight in the chttpd layer. > > 3. Mango was missing the "fields" feature but this has now been ported. > > 4. It’s important to have the latest version of couch_replicator as the > "rescan" fix is a significant performance difference. > > We’ll be working on patches for everything we’ve found and continue to find > since we now work directly against the ASF branches, but hopefully this gives > us all some confidence that the 2.0 codebase is in decent shape. Good finds! Is there an ETA on any of these, or tickets that we can follow? > I share your disappointment at the js test suite progress, it’s been over a > month and it’s not happened. Cloudant has a different test suite (in Python) > that has given us the confidence to proceed to production deployment. With my > ASF hat on we can’t consider that a factor in the couchdb release process but > it’s comfort nonetheless. That is definitely good to know, and I share your comfort. Nonetheless, at least https://github.com/apache/couchdb-chttpd/pull/98 shows that there are still some things are in 1.x that haven’t made it over to any other test suite but the JS one yet and are missing from 2.x, I want to keep this list as short as possible. > If we investigate the above issues, fix what is broken, I think we can make a > beta build even without the test suite. I hope community engagement will be > sufficient to move forward to the real 2.0 release. This has been a long > (looooonnnnnggg) time in the making and I personally very much want to look > back on the 2.0 milestone. I’m on board with this. I’d even be okay with adding the ones you mention on the CouchDB JIRA as blocking* and then call the initial release an “alpha”, and point to the JIRA blocking list as “known issues”. Subsequent releases can be called “beta” as soon as the blocking list is empty (so using “beta” in the traditional no-known-bugs meaning). * https://issues.apache.org/jira/browse/COUCHDB-2876?jql=project%20%3D%20COUCHDB%20AND%20priority%20%3D%20Blocker%20AND%20resolution%20%3D%20Unresolved%20ORDER%20BY%20key%20DESC What do others think? Also, who’s volunteering for release master? :) Best Jan -- > > B. > >> On 12 Dec 2015, at 14:40, Jan Lehnardt <j...@apache.org> wrote: >> >> Hey all, >> >> where are we on 2.0? >> >> Who is working on the 2.0 blockers listed on http://s.apache.org/couchdb-2.0? >> >> * * * >> >> I said I wanted the JS test suite to get fully passing before we get a beta >> out, but given the meagre response (thanks all who did help!), it appears >> foolish to stand firm on this. >> >> We should keep working on this, but it could be handled during the beta >> period under “known issues”. See >> https://github.com/apache/couchdb-chttpd/pull/98 for an example of just two >> tiny things we accidentally dropped between 1.x and 2.x, and I fear there >> are a lot more hidden in the 45 JS tests that we don’t run yet. Especially >> our replicator tests are part of the skipped tests and that worries me. >> >> * * * >> >> Nevertheless, we should try to get a beta out before the holidays. What is >> missing before we would feel comfortable getting a beta release out? >> >> Who can help get us over the hump? >> >> Best >> Jan >> -- >> >
