OMA
OMA copied to clipboard
microbiome
Improve Docker container
Right now Docker container is very basic and cannot be used.
I managed to run in with Podman 4.8.3 :
podman pull https://github.com/js2264/OMA/pkgs/container/OMA
podman run --rm -p 8787:8787 <container_id>
It's then accessible on localhost:8787. Username root and password is given in shell.
There are no accessible files in RStudio web-interface. Inside the container there is a OMA_0.98.16.tar.gz archive.
BiocBook supports containerization, so it a Docker image might be nice to have because it can help avoid all dependency issues on different platforms and their versions people run.
@antagomir, currently the most up-to-date container is on ghcr.io/microbiome/oma, but it's private and requires logging in with docker login ghcr.io. I just tried it on a fresh install of Fedora server and it fails at
[23/37] pages/multiassay_ordination.qmd
processing file: multiassay_ordination.qmd
|.......................................... | 74% [mofa6] Error in `py_get_attr()`:
! AttributeError: module 'mofapy2' has no attribute 'run.entry_point'
Run `reticulate::py_last_error()` for details.
Backtrace:
1. MOFA2::run_mofa(model, use_basilisk = TRUE)
2. basilisk::basiliskRun(...)
3. MOFA2 (local) fun(...)
4. MOFA2:::.run_mofa_reticulate(object, outfile, save_data)
6. reticulate:::`$.python.builtin.module`(mofa, "run.entry_point")
7. reticulate::py_get_attr(x, name, FALSE)
Quitting from lines 126-136 [mofa6] (multiassay_ordination.qmd)
Execution halted
Error in run_serve_daemon("preview", file, NULL, args, render, port, host, :
Error starting quarto
Image ghcr.io/js2264/oma:latest is outdated.
@artur-sannikov MOFA2 is causing this issue, but any idea why? Any idea how to fix this?
I'm not even able to build an image on GitHub: https://github.com/artur-sannikov/OMA/actions/runs/10055341285/job/27791808192 because it fails due to API rate limiting (it's too big, I guess). I want to try standalone Docker build without relying on BiocBook.
@antagomir, the image ghcr.io/js2264/oma:latest is not built anymore. This image should be made public. We can then close. I'll open a PR to update README.
If you can take care of closing this issue when it is ready that would be great.
Someone with access to the container registry settings, needs to make the image public.
That requires changing settings from organization level. Does it now work?
Another thing:
It seems that Docker is not working anymore since it exceeds the cache. I tried to fix it month ago (62d6764db15ddd052b4ed1199fec3bfbe1915631, d824b65de5b05ef7a939c8484febd14baf7c3525, 80a85411124dbb45bb3feab12510257d8cde0aa5, 4e80d2f1eea1a046959ad4f3a1b615d6af47b7bd, 988a78c6732402c35532dd85bd3ffd628d1e0fa2). It worked, but now it exceeds the limit again (Maybe new packages were added to OMA). See the GHA report: https://github.com/microbiome/OMA/actions/runs/10271529523/job/28421702599
Would you be able to fix this @artur-sannikov ? This is somewhat urgent since the run does not work and OMA is not updated. I would recommend involving BiocBook to this. This affects all BiocBook users and the guy might have some idea how to fix.
We can probably afford extra space when necessary but it would be anyway good to minimize the loads.
It took > 1 hour to build: https://github.com/artur-sannikov/OMA/actions/runs/10280588960/job/28448241727#step:12:8806, but the building part was successful. It was not pushed due to token permissions I think, but it's not a problem.
I think such long build time is due to miniconda installation to fix #585. I cannot build the image locally because I hit the API rate limit.