I'm on an M2 Mac and trying to build the base image from the 8305-payara6-ee10-v3 branch.
I'm seeing "Command start-domain failed" and [INFO] DOCKER> ERROR: failed to solve: process "/bin/bash -euo pipefail -c # Set admin password
Here's the output:
p6builderr.txt
Any ideas? :sweat_smile:
@Thomas van Erven since you mentioned M1/M2, lemme know if you have time to try to reproduce this. :big_smile:
I'll ask a colleague next Monday; I don't have a M2 nor have intentions to use it : )
Philip Durbin said:
Any ideas? :sweat_smile:
I thought you all were building with Ubuntu? That matches the stack trace Payara generated on fresh installs when the latest RHEL-packaged OpenJDK failed to include tzdata-java as a dependency.
Yes, the base image is Ubuntu but from my Mac (M2 at home and Intel at work), I've been running this to build the base image locally:
mvn -Pct -f modules/container-base install
It works on the intel Mac but not the M2. I'm not sure why.
I looked into your error log
Did you see that there was a Java EE OSGI error refusing to load >6.0?
2 messages were moved here from #containers > base image build matrix by Oliver Bertuch.
No, I didn't notice that.
Should we try building an arm image using GitHub Actions?
I'm not sure what you mean - we do build ARM images using GHA already?
For PRs we build app and configbaker but not base: https://github.com/IQSS/dataverse/pull/9685#issuecomment-1670371449
Oh. Yeah. True. Should we do that? Is it worth the extra minutes? We would need to push the image to some registry though - we can't keep both architectures in the local cache
We could push a P6 image manually for the time being to some place. It's not like we are altering much in the base image so far.
Yes, we should push the base image somewhere.
Not necessarily on every commit. Maybe manually with workflow dispatch?
@Oliver Bertuch is there something you'd like me to try on my M2? I have a couple hours until I bike to work.
We could try stuff if you want to
I'm around
Want me to light the fire in a Zoom room?
Sure, but gimme 5 min plz
@Oliver Bertuch https://harvard.zoom.us/j/94466543559?pwd=MC9ZVTFNT3RHd1JvbWE5V21YR2JXdz09
Now people are asking me to join a work meeting. Sry. Raincheck? (Hopefully will not take very long)
Raincheck is fine. I'll go eat breakfast. :big_smile:
Done. Should we give it a go?
Sure, but don't make me late for the ct meeting in an hour. I still need to bike to work. https://harvard.zoom.us/j/92566286460?pwd=bzdKamdrM09oYUQrdktqcmZtNTl3QT09
This seems to help!
mvn clean -f modules/container-base
My educated guess: the builder saves a state locally to reuse cached layers. Probably you got caught in the middle here, so cleaning up the target dir removed that state and you're good to go again.
It worked! Thanks, @Oliver Bertuch ! ![]()
Philip Durbin has marked this topic as resolved.
Last updated: Oct 30 2025 at 05:14 UTC