Tero Marttila

Yes, looks like `kontena-plugin-cloud` is monkeypatching `parse_response` to mangle `{"errors": [{"title": ...}]}` => `{"error": ...}`: https://github.com/kontena/kontena-plugin-cloud/blob/734b1bd1602cc5feff0059c148edd7b00f9b90c5/lib/kontena_cli_plugin.rb#L88-L102 The stack registry `errors` response doesn't have the `title`, so the mangle `error` ends...

It's different if you're tailing with `?from=` vs just querying logs. The problematic part here is the `where(:created_at.gt => since).order(:id => -1).limit(limit).to_a.reverse` - that only returns the limit most recent...

The fix for this would be to make `kontena stack rm` use the stack deploy mechanics to actually wait for the containers to terminate. The current `knotena stack rm` only...

Fix for this is in #3297, with the preparatory changes split out into #3277 and #3278 Definitely not going to be in 1.5, though

Using actual serialized YAML for the human-readable CLI output needs something to customize how `psych` dumps timestamps, I don't know how to get "nice" output without extra quoting or excessive...

Possibly related, but I'm also seeing leftover `dead` containers, which also cause duplicates once the service is re-deployed: ``` [done] Retrieving container list CONTAINER_ID IMAGE COMMAND CREATED STATUS localhost/sleep.sleep-1 alpine:3.2...

Initial guess here is that the agent [`EventWorker#publish_event` -> `rpc_client.async.request('/containers/event', ...)`](https://github.com/kontena/kontena/blob/3d608c58a0c0ce046e3a3f2065c39431b2fbc13f/agent/lib/kontena/workers/event_worker.rb#L86) for the `stop`, `die`, `destroy` events is racing with the [`ContainerInfoWorker#publish_info` -> `rpc_client.async.request('/containers/save', [data])`](https://github.com/kontena/kontena/blob/3d608c58a0c0ce046e3a3f2065c39431b2fbc13f/agent/lib/kontena/workers/container_info_worker.rb#L76) such that the delayed `/container/save`...

Fix for the race between the two `EventWorker` and `ContainerInfoWorker` actor RPCs would be to only have one actor responsible for sending the RPCs, and making sure they happen in...

Removing from the 1.5 milestone because the workaround for this is to run the e2e specs using `gem build kontena-cli.gemspec && gem install kontena-cli-*.gem` to get the standard rubygems binwrapper.

Disabled the broken e2e specs for now: #3359 This issue remains open because the specs are still broken, and needs more investigation into why.