I'm building my image for Durable %SYS with Docker volumes:

FROM intersystemsdc/irishealth-community:2023.1.0.229.0-zpm
#FROM containers.intersystems.com/intersystems/irishealth-community:2023.1.0.229.0


USER root
RUN mkdir -p /iris/durable && chown ${ISC_PACKAGE_MGRUSER}:${ISC_PACKAGE_IRISGROUP} /iris/durable
USER ${ISC_PACKAGE_MGRUSER}

And start it like this:

version: "2.2"
services:
  iris:
    user: "51773:51773"
    build: 
      context: .
      dockerfile: Dockerfile
    command: --check-caps false
    volumes:
      - iris-durable:/iris
    environment:
      - ISC_DATA_DIRECTORY=/iris/durable
volumes:
  iris-durable:

If I use the base image containers.intersystems.com/intersystems/irishealth-community:2023.1.0.229.0 it works as expected, but if I use intersystemsdc/irishealth-community:2023.1.0.229.0-zpm it fails with:

iris  | [INFO] Executing command /docker-entrypoint.sh iris-after-start ...
iris  | [ERROR] /docker-entrypoint.sh: line 188: /usr/irissys/iris.init: Permission denied
iris  | [ERROR] Command "/docker-entrypoint.sh iris-after-start " exited with status 256
iris  | [FATAL] Error executing post-startup command

when containers needs restart or stop, it skips shutdown iris procedure

why using an underscore ? IRISUSERNAME and IRISPASSWORD are env used by embedded python, using "_" underscore can add confusion 

Why does the ZPM image add 400 MB?

I am just getting up to speed with docker and am having trouble pinpointing why a simple library is taking up so much space.

If you know the answer to this, perhaps you could also explain why the irishealth image is 2GB larger than the vanilla iris image?

                Created4.4s 1  Creating                                                                                      6.5s

[+] Running 2/3m-dockerhub-iris-1
✔ Network zpm-dockerhub_default
Created4.4s atform (linux/arm64/v8) does not match the detected host platform (linux/amd64/v2) and no specific pla
[+] Running 2/3m-dockerhub-iris-1
✔ Network zpm-dockerhub_default

intersystemsdc/iris-ml-community:2022.3.0.599.0-zpm is built on top of an unreleased build (599). The official community edition build for 2022.3 is 606. If users try and use the 2022.3.0.599.0-zpm tag, they will see this error. It does appear that latest now points to latest: 2022.3.0.606.0, which is good and should allow users to get the correct base build. However, does it make sense to remove the 599 build so that users do not get errors when trying to use it?

For context, if you try and train a model on the 599 build, you see this error:
[SQLCODE: <-400>:]

[%msg: <ERROR #5002: ObjectScript error: %OnInit+20^%ML.AutoML.Provider.1^1^ *<class 'pkg_resources.extern.packaging.version.InvalidVersion'>: Invalid version: '1.1build1' - Import>]

Hi @evshvarov

I'm seeing this in the docker-compose logs:

iris_1  | 11/13/20-04:26:41:489 (384) 2 [Utility.Event] Preserving journal files /usr/irissys/mgr/journal/20201103.001 and later for journal recovery and transaction rollback

That date - 2020-11-03 - appears to be when this image was created. I couldn't determine what is wrong, but did confirm this doesn't happen in the store/intersystems... image without zpm.

Do we need to fix this image or do you know of way I can clear need for journal recovery?