Juju Team Updates - Week #25 2021

Hi everyone, below you can find the updates from the Canonical Juju Team for the week starting on the 21st of June 2021.

Have a great weekend!

Working on a PR to fix bug 1928778 - juju refresh not working when upgrading from a pod-spec to sidecar charm.



I have been working on Juju 2.8 -> 2.9 upgrade issues.

  • Landed lp1930798
  • Fixed a namspace discovery issue that was causing model operator’s in existing models to be re-instantiated
  • Almost finished last upgrade bug for old Kubernetes credentials that need to be backwards compatible in 2.8 models on a 2.9 controller.

I have been working on a fix for LP1932547 - juju ssh --proxy not working on aws when targeting containers with FAN addresses.

In addition, I completed the spec for introducing an assumes section in charm V2 metadata.


I have fixed a few issues:

  • migrated CaaS model cannot be destroyed - lp-1927656;
  • juju ssh is broken for CaaS controller models - lp-1929904;
1 Like

In the last week and a bit I’ve worked on:

  • 2.9.6: Make sidecar unit IP address visible immediately. #13092 LP1929364
  • 2.9.7: Add support to ec2 provider for encrypted root disks, kms, gp3, io2, throughput. #13101 LP1921139

Continuing work on Pebble, currently looking into making service Start/Stop/Autostart all idempotent. Exploring adding a new method that reconciles the running services with the plan, by restarting services that have changed in addition to Autostarts current behaviour.

1 Like
  • 2.9: Closing out work on the resilience for charmhub retries: lp-1928182
  • 2.9: Metrics for outbound API connections (currently only charmhub) - driveby for lp-1928182
  • 2.9: Improved logging UX, by using labels to group logs


It will soon be possible (2.9.7 release) to see charmhub outbound API requests from the controller. Using the following documentation: Collecting Juju Metrics, you can wire up prometheus to graph the data.

View the metrics of a localhost (lxd) setup using the following curl command:

curl -k -s https://<IP>:17070/introspection/metrics -u user-prometheus:<password> | grep outbound

The metrics are lazily created, so a juju find, juju info, juju deploy ubuntu is required.


Labels for logging will allow the aggregation of logs via a label rather than the namespace.

juju model-config logging-config="#http=TRACE"

The above will turn on HTTP loggers to trace. This is a new UX feature to help with debugging, it’s not been full worked through Juju yet and might be subject to change.

1 Like

Notable from the week:

  • Wrangling the 2.9.5 official and 2.9.6 candidate releases.
  • Fixing LXD bugs and landing the support for LXD projects.
1 Like

Merged a fix for LP:1928796 Unable to deploy bundle with sidecar and pod_spec charms.

Working on a fix for a panic in getting MongoInfo from an agent config and how we got to the state where an agent.conf had no addresses.

Investigated to handle the unit upgrade step in juju 2.8, if moving from juju < 2.8 to juju 2.9, as the step is skipped. Juju 2.9 has no unit agents.

1 Like