build faliure - mondo-minimal

[harryhoch]

Seems to blow up when trying to pull in mondo-minimal

• Deployment error
• GitHub Actions script
• Makefile

Mondo Release

[harryhoch]

@LucieContamin , I think i have temporarily fixed this by pointing the build process directly at an older version of the midas-minimal.owl file. Please run the action and take a look to see if it looks ok. I'm trying to get an answer from Bill Hogan as to whether there's a more durable fix.

[LucieContamin]

It seems that the GitHub Action has already re-run with success and the last version with the license has been deployed. I am afraid that if I trigger the Action again it will create a new release without any change

[harryhoch]

Ok. that’s fine. I’m going to continue to look for a more permanent fix.

…

-harry From: Lucie Contamin ***@***.***> Date: Wednesday, November 6, 2024 at 8:34 AM To: midas-network/midas-data ***@***.***> Cc: Hochheiser, Harry ***@***.***>, Author ***@***.***> Subject: Re: [midas-network/midas-data] build faliure - mondo-minimal (Issue #67) It seems that the GitHub Action has already re-run with success and the last version with the license has been deployed. I am afraid that if I trigger the Action again it will create a new release without any change — Reply to this email directly, view it on GitHub<#67 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AABKU43ZTN3O7LQ6VR6X3FDZ7ILHTAVCNFSM6AAAAABRHICKAWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJZG43DSMRTG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[harryhoch]

From @JohnLevander:

Hi Harry,

A few updates on this issue..

1. I updated the version of robot.jar to the 1.9.7 from 1.9.4. No NEW errors, but still we run out of heap space. I'd say we probably want to use 1.9.7 moving forward.

2. Free Linux machines only have 16GB ram and 14GB SSD. Even if the Java instance got all the ram and virtual memory, that's only 30G of memory that we could ever use on the free plan.

3. They sell "-xl" runners, e.g. ubuntu-latest-xl, which gives you much more memory and space (up to 256GB RAM and 2040GB SSD), but obviously there is a fee and I'm sure we are trying to avoid that. I tried running using an -xl runner 20 hours ago, and I still see a message "Waiting for a runner to pick up this job..." so it doesn't seem like there is any free support for xl machines.

Plan:

I think the best solution is to get mondo to include mondo-minimal in their release. If they do that, we are done messing with this issue. I filed an issue with their team here: monarch-initiative/mondo#7912.

If they can't/won't include mondo-minimal in their release, I also asked them if they could provide guidance on how to "trim" down the release ontology to make it contain only the "minimal" entries. Hopefully it won't be that hard if it comes down to that.

I have other ideas, but let's see what mondo says first. In the meantime if we want to create a release, we should be able to do it locally: https://www.browserstack.com/guide/test-github-actions-locally

[harryhoch]

Thanks, @JohnLevander. Note that it's working ok for now based simply on pulling back to the last released version of mondo-minimal.

A couple of points:

1. The robot conversion seems almost to be a null step. If we skip that, would it work?
2. Given the makefile, it's easy enough to run locally without doing much.

I'm going to try to look into #1.

[matentzn]

Hello @harryhoch and @JohnLevander!

I made a PR here #68 that replaces mondo minimal by "simple", which is very close to the old (now deprecated) minimal subset. Some additional observations:

1. Mondo simple certainly should not cause an out of memory error even in the simple ubuntu image.. We have hundreds of actions using Mondo in that very image!
2. Re ROBOT, note that the release action runs on ODK - so you should update that to ODK 1.5.3 to get the latest (or almost latest) ROBOT (not install some local ROBOT version). You updated ROBOT locally to the latest version I assume, which means you probably had a memory error because you didn't assign enough memory in ROBOT_JAVA_ARG?

[harryhoch]

ok thanks. I think this is working.