Machine cloud-init takes a long time when one agent binary URL times out
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Canonical Juju |
Fix Released
|
High
|
Andrew Wilkins |
Bug Description
When machine agents are deployed to separate private networks from the controller, e.g. as in any Azure controller, the machine agents started in non-controller models will be slow to download the agents.
The reason for this is that the cloud-init script attempts to download from each of the controller's addresses in series, starting with the cloud-local addresses. These are never usable by machine agents except for those within the same model.
We should do at least one of the following things:
1. attempt to download the agent from each of the addresses in parallel
2. specify a shorter timeout to curl
3. update providers to ignore addresses they know they can never use (e.g. the cloud-local address for azure, except when in the controller model)
summary: |
- Machine cloud-init hould attempt to download agent using all addresses + Machine cloud-init should attempt to download agent using all addresses in parallel |
Changed in juju: | |
milestone: | 2.0-rc1 → 2.0-rc2 |
Changed in juju: | |
milestone: | 2.0-rc2 → none |
Changed in juju: | |
milestone: | none → 2.0.0 |
Changed in juju: | |
milestone: | 2.0.0 → 2.1.0 |
Changed in juju: | |
status: | Triaged → In Progress |
Changed in juju: | |
status: | In Progress → Fix Committed |
Changed in juju: | |
status: | Fix Committed → Fix Released |
PR against 2.1: https:/ /github. com/juju/ juju/pull/ 6973