juju-solutions / build-cloud Goto Github PK

It would be nice to say "this test plan should run on aws and gce". We know, for example, that bigtop big data bundles do not work on joyent, so we would like to blacklist that cloud at a test-plan level.

If not specified, cwr should assume all clouds.

cwr.jujucharms.com hasn't had a good run on any cloud for about 2 weeks on various bundles. We either see a blank (no run) or yellow triangle (infrastructure issue). Per our local testing, we expect to see greens on the major clouds (AWS, Azure, and GCE) for kubernetes-core, hadoop-processing, hadoop-spark, and hadoop-kafka.

Please let us know when cwr.jujucharms.com is back to a healthy state.

-thanks,
Antonio

I haven't seen any successful runs on OpenStack cloud provider. This bug is to track getting repeatable successful runs on the OpenStack Cloud provider.

Add "url" key in the test plan and display the link to the bundle on the report page.

When a bundle has a test failure submit a bug via the upstream bug link.

We should only runs bundles that can reliable run across all clouds. The bundle should not be tuned to the OB. The bundle should be the upstream bundle that works on the OB.

Looking at http://cwr.jujucharms.com/bundle_hadoop_processing/index.html we haven't seen consistent AWS runs. We saw one success on Jan 6, but haven't seen a reliable run since Dec 11.

I haven't seen any successful runs on the MAAS cloud. I have seen some runs report, but most often they are flagged by Infrastructure issue.

This bug is to track making MAAS a successful repeatable cloud substrate, inline with the public clouds like AWS, GCE, and Azure.

Add a footer to the bottom of cwr.jujucharms.com to submit and contriute.

May also be good to have a message somewhere on the page that says if you would like to run CWR yourselves there is a CWR tool at https://github.com/juju-solutions/cloud-weather-report

Also while on bugs we should look at moving this repo and cloud-weather-repot to github.com/juju.

We are consistently seeing GCE fail due to infrastructure issues due to, "Unable to upload scripts."

2017-01-03 09:52:33 Deploying applications...
2017-01-03 09:53:44 Config specifies num units for subordinate: ganglia-node
2017-01-03 09:54:48 Config specifies num units for subordinate: plugin
2017-01-03 09:55:53 Config specifies num units for subordinate: rsyslog-forwarder-ha
2017-01-03 09:55:57 Adding relations...
2017-01-03 09:55:57 Deployment complete in 205.80 seconds
b'ERROR cannot connect to any address: [104.199.15.106:22 10.240.0.8:22]\n'
b'ERROR cannot connect to any address: [104.199.15.106:22 10.240.0.8:22]\n'
Unable to upload scripts

An example of this is in http://cwr.jujucharms.com/bundle_hadoop_processing/1d62d1d69dbc439eb45edca11d80cfac/report.html

• for 01-bundle.py

But consistently for GCE for hadoop-processing.

This bug is to track the prefix of "bundle:" to bundle names. Suggest to align where we do not prefix with the bundle name.