AndreyByhalenko
commented
1 year ago Hello, Do you have any estimated for the solution of this issue? Thanks
Open Clare-Octopus opened 2 years ago
AndreyByhalenko
commented
1 year ago Hello, Do you have any estimated for the solution of this issue? Thanks
michelle-luana
commented
1 year ago Hi Andrey At this stage this isn't actively being worked on as we haven't had reports from other customers being impacted. We'll continue to monitor and if more customers are impacted by this we will reevaluate. Thanks Michelle
sluebbert
commented
1 year ago We had the same issue. We were using robot accounts in Harbor to use in Octopus Deploy. Harbor defaults to the robot username prefix of robot$. While this works in the Octopus web app just fine, it does not work at the time of deploying a release.
Harbor logs at that moment showed:
failed to authenticate user:robot, error:Failed to authenticate user, due to error 'Not found an entry'
Since the user we originally used for Octopus was robot$octopusdeploy, and the user logged by Harbor was just robot, we assumed Octopus wasn't escaping $ very well or just didn't support that char in some way for Docker usernames.
Changing the robot name prefix in Harbor to something like robot_ instead resolves the issue. Just be careful as that seems to immediately rename all of your robot account names in Harbor when you change it.
The UI searching auth errors you are referring to as well probably relates to how Harbor robot accounts do not have permission to query the registry catalog unless set as admin accounts or unless you set it through the Harbor API at the time of writing this. https://github.com/goharbor/harbor/issues/6784 https://github.com/goharbor/community/pull/223
Ignore all of this if you aren't using a robot account :)
kengibous
commented
9 months ago this is indeed still an issue - i believe certain characters are not being escaped correctly such as $
Team
Severity
Blocking customers that use Harbor Feeds from deploying docker images.
Version
2022.2.6764
Latest Version
I could reproduce the problem in the latest build
What happened?
When deploying a docker image using a 'Run a Docker Container' step it errors out with login failures.
The feed tests successfully in external feeds and will pull a package:
One thing I did notice was if you don't search for the full package you get this error:
The step also shows the same error if you just search for a partial package name. I captured the error in a HAR File placed here (Internal Link - Under clare.martin@octopus.com and name Harbor)
It does work if you type the full package name though:
Two customers (same forum post) have reported the same issue (slightly different initial error messages but both login failures and both feeds are testing fine and finding packages).
Reproduction
Error and Stacktrace
More Information
Errrors both users are getting are slightly different -
One user suggests - It seems like the credentials are not getting passed over to the tentacle. I have even tried manually using “docker login” on both base and root user to have the docker daemon already logged into the registry with no luck.
From what I can see both users have IP addresses for their feeds - Harbor is only On-Prem, they do not do a cloud version, so the difference in error messages to mine might just be because of the URL differences (my test was done on cloud, theirs on prem).
Workaround
None, login details are correct as external feed authenticates and pulls the packages on a test.