| cjp256 | holmanb: i noticed that with the ds_detect changes that the datasource no longer will fall back to None as it had with is_platform_viable() check in get_data(). Specifically in the case of Azure, the ds-identify or user ds config would be [Azure, None] and get_data would bail if detection failed with something like "No local datasource found". End result is customers using that config on non-Azure platforms now end up | 18:09 |
|---|---|---|
| cjp256 | executing the _get_data() path as if on Azure, causing some issues | 18:09 |
| cjp256 | I notice that ds-identify detection for Azure is broader than the ds_detect method, which I think we need to unify too | 18:11 |
| holmanb | cjp256: agreed on ds-identify / ds_detect disparity - those should match | 18:42 |
| holmanb | cjp256: iirc checking datasource_list in override_ds_detect was proposed during review as an optimization do avoid re-checking what ds-identify already checked - in that case [Azure, None] is expected I think | 18:44 |
| holmanb | cjp256: if [Azure, None], I believe it will still fall back to None whenever get_data() times out | 18:46 |
| holmanb | *to avoid re-checking | 18:46 |
| holmanb | cjp256: When customer provides [Azure, None] and cloud-init runs _get_data() is this just a change in expected behavior in a failure path? Or is fallback to DataSourceNone a normal behavior in user workflow? | 18:52 |
| cjp256 | normal behavior if image is used outside Azure | 18:52 |
| cjp256 | right now that timeout is a pretty long wait so ideally we’d like to go back to skipping that path if detection determines it won’t (shouldn’t) work | 18:54 |
| cjp256 | holmanb: what do you think about https://github.com/canonical/cloud-init/pull/4426 | 21:09 |
| holmanb | cjp256: seems reasonable to me, I think | 21:45 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!