Join GitHub today
GitHub is home to over 20 million developers working together to host and review code, manage projects, and build software together.
can't deploy hadoop-processing with python-libjuju #67
Comments
|
As @kwmonroe mentioned, I am going to pick this one up, as it breaks matrix. |
|
This basic test case worked fine for me: async def main():
model = Model()
await model.connect_current()
try:
await model.deploy('cs:hadoop-processing')
finally:
await model.disconnect()
loop.run(main()) |
|
@johnsca Intriguing. I bet that there's something broken in the local bundle parsing code then. |
|
I cut the original juju status short.. Check this out: It's missing some of the "rsyslog-forwarder-ha" relations. Here's the bundle it used: http://paste.ubuntu.com/24096487/ It's like it stopped on line 80 and skipped 81-84. So maybe matrix is waiting on all the relations to be added before proceeding to add-unit. Edit, you can see from the matrix.log that it did indeed stop adding relations after rsyslog-fwrd::client: |
|
@petevg I tried again with a local copy of the @kwmonroe The from juju import loop
from juju.model import Model
async def main():
model = Model()
await model.connect_current()
try:
await model.deploy('/tmp/hp')
finally:
await model.disconnect()
loop.run(main()) |
|
@johnsca I tried your test from within my cwr/0 unit, lxd exec'd into the cwr container that previously failed. It worked :/ Not sure why the original stopped processing |
|
I'm seeing intermittent issues similar to this with conjure-up. I think there's some sort of race condition for this. |
|
I'm currently working on this. I have a fix locally, but I'm going to check in a more complete fix that also addresses #65 |
petevg
referenced this issue
Mar 7, 2017
Closed
We don't interpret all errors from the api as errors #65
|
PR here: #74 |
kwmonroe commentedMar 2, 2017
•
Edited 1 time
-
kwmonroe
Mar 2, 2017
I recently ran cwr-ci on https://jujucharms.com/hadoop-processing/. CWR calls bundletester which calls matrix, which does libjuju stuff.
I think it adds apps, machines, and then adds units -- it feels like the add-unit piece isn't working. A matrix model was created, apps were added, machines were added, but no units ever showed up:
$ juju status -m matrix-close-dingo Thu Mar 2 15:26:05 2017 Model Controller Cloud/Region Version matrix-close-dingo lxd localhost/localhost 2.0.3 App Version Status Scale Charm Store Rev OS Notes client blocked 0 hadoop-client jujucharms 3 ubuntu ganglia waiting 0 ganglia jujucharms 5 ubuntu ganglia-node waiting 0 ganglia-node jujucharms 6 ubuntu namenode maintenance 0 hadoop-namenode jujucharms 8 ubuntu plugin maintenance 0 hadoop-plugin jujucharms 8 ubuntu resourcemanager maintenance 0 hadoop-resourcemanager jujucharms 8 ubuntu rsyslog waiting 0 rsyslog jujucharms 7 ubuntu rsyslog-forwarder-ha waiting 0 rsyslog-forwarder-ha jujucharms 7 ubuntu slave maintenance 0 hadoop-slave jujucharms 8 ubuntu Unit Workload Agent Machine Public address Ports Message ### HEY! THERE'S NOTHING HERE Machine State DNS Inst id Series AZ 0 started 10.38.19.117 juju-d91d66-0 xenial 1 started 10.38.19.56 juju-d91d66-1 xenial 2 started 10.38.19.225 juju-d91d66-2 xenial 3 started 10.38.19.79 juju-d91d66-3 xenial 4 started 10.38.19.119 juju-d91d66-4 xenial 5 started 10.38.19.166 juju-d91d66-5 xenial Relation Provides Consumes Type juju-info client ganglia-node subordinate hadoop-plugin client plugin subordinate juju-info client rsyslog-forwarder-ha subordinate node ganglia ganglia-node regular ...I was able to manually go in and add units like this:
But I'd like to not have to do that. @petevg is on the case!