Docker setup #502
Docker setup #502
Conversation
Codecov Report
@@ Coverage Diff @@
## develop #502 +/- ##
===========================================
- Coverage 77.71% 77.69% -0.02%
===========================================
Files 61 61
Lines 5080 5080
Branches 697 697
===========================================
- Hits 3948 3947 -1
- Misses 922 923 +1
Partials 210 210
Continue to review full report at Codecov.
|
|
Are you seeking a review on this still? If so, I suspect @mimming would be your best bet. |
I might change this still so am leaving it open. I'm not sure that starting up the jupyterlab is the best default option. |
* Reducing size * Starting CMD correctly * Skipping git for zip
* Adding env variable for solve-field * Adding git back to included * Start up a bash shell for CMD Note: It's expected this is used as a base for other images and so the CMD will be overriden.
|
|
||
| #### Let Docker use gcloud | ||
|
|
||
| Docker needs to be able to use your `gcloud` login to pull the PANOPTES images. There |
There was a problem hiding this comment.
Do you think that the "advanced-authentication" is still necessary? The page now recommends just using this command:
gcloud auth configure-docker
After that (assuming I'd not previously installed docker-credential-gcr, which it doesn't appear that I did according to gcloud components list), running this command worked:
docker pull gcr.io/panoptes-survey/pocs:latest
There was a problem hiding this comment.
This could have changed, I'll check it out.
| docker run -it -p 9000:9000 --name pocs gcr.io/panoptes-survey/pocs | ||
| ``` | ||
|
|
||
| The POCS image will automatically start [Jupyter Lab](https://jupyter.org/) running |
There was a problem hiding this comment.
This appears to be aspirational, but not currently true. The command is just /bin/bash.
There was a problem hiding this comment.
Actually it worked perfectly fine but I removed it. If we leave the CMD as /bin/bash then you can launch whatever you want when you start the instance so it's a bit more generic.
I'll fix the README, thanks.
|
Looks like ffmpeg needs to be added to the deps used when creating the image. |
|
And now I see that the same thing (ffmpeg being missing) happens in travis-ci. |
Will do.
It could certainly be optimized some more. Also, this is acting as a lower layer for the piaa and other docker containers that are in there. One thing to keep in mind is that my update of this Docker changed the scope a bit although it might not matter. I was originally making this for the running of a unit and/or jupyter-lab instance connected to the unit. The only way I'm currently using it as a base to build an image for compute instances for doing algorithm work. However, when I start those instances I want solve-field, sextractor, etc. without having to install. Layers will help but we should also figure out exactly how we want to use the containers.
I see that occasionally (it's always there) and keep forgetting to do anything about it. Should probably just suppress warnings as we don't actually need to make the movies on travis. |
|
@jamessynge I think this could be reasonably merged. Could feasibly remove the changes to the README and just leave the Dockerfile. Thoughts? |
|
How about creating a docker directory, and putting the Dockerfile and relevant portions of the README into that directory. Then I think this would be good to go. |
|
I want to build on this, so rather than wait for your updates, I'm going to merge as is, then will make changes I suggested. |
Adding initial Dockerfile and updated instructions.
We could design the container better so it is not such a monolithic service but this gets at least the initial file there. Also needs to be reduced in size.
Closes #27