Closed Remi-Gau closed 1 year ago
Persisting the image is taking too long it looks like. Might need to see if we can streamline the docker file.
Looks like the image
builds
but the workspace
for the tests
runs out. CI
suggests to use parallelism
for tests
. I have never done that, anyone has an idea how to implement this?
other option: we do not persist anything to the workspace and just only ever get stuff from the cache in the follow up steps?
Sorry, it's building the image that's taking 42 minutes. I'll try a large or x-large machine to see if that speeds things up. If that doesn't work. We could download freesurfer in a previous step and persist it, then copy it into the container instead of downloading it. That might speed up the bulid, but it'd require some editing of the dockerfiles.
we may want to check if switching to github action should be considered.
It's not clear to me why the Freesurfer 7 build is now taking half an hour. It was taking ~16 minutes on previous commits and I don't see any changes in the dockerfile in the mean time. @Remi-Gau switching over to github actions would be one option, but I don't have time to do an overhaul like that right now.
@PeerHerholz, I don't think the length of time it's taking to build is something that parallelism is likely to solve.
It looks like the large machine might just barely sneak through. If that fails I'll try an x-large machine and see if it goes any faster, but I think we might need to cache the freesurfer tar if we want to get a more reliable build time.
Ah, so there are intermittent timeout failures in both the build step and in the test step. Those test failures could maybe be fixed with parallel, or by splitting the tests into two steps.
On Thu, Sep 7, 2023 at 2:02 PM Remi Gau @.***> wrote:
Something is indeed strange.
2 circle CI runs on the same commit https://github.com/bids-apps/freesurfer/commit/466355952a4a32fd1a11f7db6d56ebbfe302baf5 :
- one fails: https://app.circleci.com/pipelines/github/bids-apps/freesurfer/111
- one passes: https://app.circleci.com/pipelines/github/bids-apps/freesurfer/106
— Reply to this email directly, view it on GitHub https://github.com/bids-apps/freesurfer/issues/78#issuecomment-1710570602, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABILLGFO4RCFIMDAN3V7S4TXZID4TANCNFSM6AAAAAA4H5CW3A . You are receiving this because you commented.Message ID: @.***>
I've fixed the build timeouts for now by skipping the cacheing step in #77.
The test timeouts are going to be a bit trickier and should be saved for a separate PR.
@Shotgunosine If you are not going to work on it right now, I may PR something to make a push to main triggers a push of "unstable" tag version to docker hub.
See an example of another app here: https://hub.docker.com/r/bids/antscorticalthickness/tags
This allows users to try the "bleeding" edge of the app.
Does this not already do that? https://github.com/bids-apps/freesurfer/blob/62eb27608575ff403a78f224b71102d73fadcc7f/.circleci/config.yml#L205-L207
ah yes sorry I missed that
But it seems that there is no deploy planned on this push to master
I suspect it is because we are ignoring deploys on all branches:
and we only do it on tags
Gotcha, I’m not planning on working on this anymore today, feel free.
On Thu, Sep 7, 2023 at 4:06 PM Remi Gau @.***> wrote:
I suspect it is because we are ignoring deploys on all branches:
and we only do it on tags
— Reply to this email directly, view it on GitHub https://github.com/bids-apps/freesurfer/issues/78#issuecomment-1710707277, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABILLGA7IESC62SKH2M6ZJ3XZISOHANCNFSM6AAAAAA4H5CW3A . You are receiving this because you were mentioned.Message ID: @.***>
https://app.circleci.com/pipelines/github/bids-apps/freesurfer/111/workflows/00c61725-eefe-4eb6-a748-c4b3586ce2fe/jobs/445