Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Paul Capestany
    @PaulCapestany
    yeah, same issue even when using {{ .COUCHBASE_SERVER_IP }} https://gist.github.com/PaulCapestany/b58b26d44b4cb6fd1f34
    Paul Capestany
    @PaulCapestany
    @tleyden I think I may have tracked things down a bit further → https://github.com/couchbaselabs/couchbase-server-coreos/issues/18#issuecomment-106026386
    Traun Leyden
    @tleyden
    just posted to the ticket
    Traun Leyden
    @tleyden
    @PaulCapestany merged. glad you remembered to run go generate!
    to use that change, you'll need to follow the instructions here: https://github.com/tleyden/couchbase-cluster-go#running-on-the-latest-code
    Paul Capestany
    @PaulCapestany
    yeah
    Traun Leyden
    @tleyden
    (so that it pulls the latest couchbase-cluster-go rather than using the version baked into the docker image)
    Paul Capestany
    @PaulCapestany
    one thing I noticed though is that go generate output seems to differ a bit (not just in generated bytes), for example removing a ~ ... does that matter?
    couchbase-cluster-go (Git) 2015-05-27 14-11-19.png
    Paul Capestany
    @PaulCapestany
    @tleyden I think etcdctl set /couchbase.com/enable-code-refresh true may not be working as expected since I just spun up new cluster and am still getting Error reading config file sync_gateway: open sync_gateway: no such file or directory -- rest.ParseCommandLine() at config.go:294
    Traun Leyden
    @tleyden

    example removing a ~ ... does that matter?

    I don't think so

    still getting Error reading config

    which command did you use to kick off the cluster? also can you paste the logs from the "cluster kick off" command?

    Paul Capestany
    @PaulCapestany
    I set up cluster with:
    etcdctl set /couchbase.com/enable-code-refresh true && \
    sudo docker run --net=host tleyden5iwx/couchbase-cluster-go couchbase-fleet launch-cbs \
      --version latest \
      --num-nodes 3 \
      --userpass "user:passw0rd" \
    && \
    sudo docker run --net=host tleyden5iwx/couchbase-cluster-go sync-gw-cluster launch-sgw \
      --num-nodes=1 \
      --config-url=https://gist.githubusercontent.com/PaulCapestany/d24c9ca8223faab66b25/raw/638729f249f65dd6a275b67d467ac46938e9005e/pac_sg_config.json \
      --create-bucket todos \
      --create-bucket-size 512 \
      --create-bucket-replicas 1
    interestingly, those logs show 2015/05/27 21:43:26 Update-Wrapper: updating to latest code
    so it looks like update-wrapper.go is trying to do its thing, logs show new downloads being triggered too..
    Paul Capestany
    @PaulCapestany
    oops, logs I'd posted above were from journalctl, here's the actual output from running the cluster kickoff commands: https://gist.github.com/PaulCapestany/35065c2fa242291a5015
    Traun Leyden
    @tleyden
    crap, it's not spitting out the sync gw fleet units.
    can you run fleetctl cat on the sync gateway units and paste the output? was this a "fresh cluster"? I wonder if the old unit files never got overwritten?
    could you try running https://github.com/tleyden/couchbase-cluster-go#destroying-the-cluster first and then trying again?
    btw it's possible to get into the container and see the couchbase-cluster-go commit id via:
    $ docker exec -ti <container_id> bash
    # cd /opt/go/src/github.com/tleyden/couchbase-cluster-go
    # git status
    Paul Capestany
    @PaulCapestany
    yeah i just came to same conclusion about fleet units
    they don't get updated it seems
    still show old ones
    (and yeah, it was a totally fresh cluster)
    Paul Capestany
    @PaulCapestany
    and yeah... git status shows that couchbase-cluster-go is on freshest commit. seems like fleet unit files just don't get swapped out with current enable-code-refresh implementation
    Traun Leyden
    @tleyden
    we need to have it spit out the sync gateway unit files into the logs so we can see that it's generating the correct unit files
    Paul Capestany
    @PaulCapestany
    would it make sense to use an automatic hook to deploy new docker images when there are commits to the couchbase-cluster-go repo? cuz that's what you're basically trying to do with enable-code-refresh, or no?
    Traun Leyden
    @tleyden
    yes and no, enable-code-refresh should have a much tighter loop since it doesn't have to wait for a docker image rebuilt (can take up to 30 mins)
    I'm making that logging change I mentioned right now
    Paul Capestany
    @PaulCapestany
    gotcha
    Traun Leyden
    @tleyden
    ok just pushed to master
    Paul Capestany
    @PaulCapestany
    yup, gonna try on new cluster
    Paul Capestany
    @PaulCapestany
    @tleyden the log statements you added don't even get called
    like, I see the code that does go get -u -v github.com/tleyden/couchbase-cluster-go/... ...which should install new binary, right? but for some reason the new binary isn't being used?
    Traun Leyden
    @tleyden

    you're gonna hate me for saying this, but "works on my computer"

    I just ran it and:

    https://gist.github.com/tleyden/005c62266ea5e87f7810

    that wasn't a "fresh cluster" .. I had a cluster running (CoreOS alpha 681.0.0 btw), and ran the commands under https://github.com/tleyden/couchbase-cluster-go#destroying-the-cluster to clean out previous fleet units and other residue (wiped /opt/couchbase/var)
    Paul Capestany
    @PaulCapestany
    ok, I've just been running as a fresh cluster each time
    Traun Leyden
    @tleyden
    hmm. I mean it should work there too.
    Paul Capestany
    @PaulCapestany
    ha. just realized I wasn't using the "update-wrapper" argument when trying to use enable-code-refresh. doh
    of course it was something really dumb
    Traun Leyden
    @tleyden
    ohhhh
    that would explain things
    Paul Capestany
    @PaulCapestany
    yeah
    trying again... with the actual proper args...
    Traun Leyden
    @tleyden
    the enable-code-refresh just sticks a value into etcd. but the command needs to be wrapped in update-wrapper to look at the value in etcd and act on it. (it rebuilds the code, then calls the actual stuff you want to call)
    Paul Capestany
    @PaulCapestany
    yeah, the sad thing is I knew that. just hadn't noticed i was missing the crucial "update-wrapper" part
    Traun Leyden
    @tleyden
    haha. well now you'll never forget :P
    Paul Capestany
    @PaulCapestany
    seriously.
    Patrick Heneise
    @PatrickHeneise
    @tleyden going to try the couchbase container on google cloud these days. Will let you know if it works.
    Traun Leyden
    @tleyden
    @PatrickHeneise glad to hear. You might want to check out https://github.com/couchbase/kubernetes which has been tested on GKE