The mirror workflow now uploads new OTA images to GCS, after which the extraction workflow downloads them again. It might be cheaper to upload the images as workflow artifacts and let the extraction workflow retrieve them from there if they are available and otherwise fall back to the GCS mirror. Running them on the same machine was also suggested, but it makes us less flexible because the extraction must always run on macOS13 images.
As mentioned here: https://github.com/getsentry/symx/issues/13, the download from GCS is the least costly part of the extraction workflow, so I leave this only here as something that came up during discussions but should only be tackled after we resolved other bottlenecks.
The mirror workflow now uploads new OTA images to GCS, after which the extraction workflow downloads them again. It might be cheaper to upload the images as workflow artifacts and let the extraction workflow retrieve them from there if they are available and otherwise fall back to the GCS mirror. Running them on the same machine was also suggested, but it makes us less flexible because the extraction must always run on macOS13 images.
As mentioned here: https://github.com/getsentry/symx/issues/13, the download from GCS is the least costly part of the extraction workflow, so I leave this only here as something that came up during discussions but should only be tackled after we resolved other bottlenecks.