There is something wrong with the recent submitted save code now requests in production: associated tasks are marked as scheduled since at least a week and should already have been executed (see https://archive.softwareheritage.org/browse/origin/save/#requests).
At the time I wrote the task, I thought that what I observed locally was related to that issue but it turns out that my configuration was outdated
I also observed the same behavior in my local development environment, I can schedule tasks but my celery worker will never receive the messages for executing them.
For the record, my worker configuration is the following:
scheduler: cls: remote args: url: http://localhost:5008/ celery: task_broker: amqp://guest@localhost// task_modules: - swh.loader.git.tasks - swh.loader.mercurial.tasks - swh.loader.svn.tasks - swh.deposit.loader.tasks - swh.vault.cooking_tasks - swh.loader.tar.tasks - swh.loader.pypi.tasks task_queues: - swh_loader_git - swh_loader_svn - swh_loader_tar - swh_loader_mercurial - swh_loader_svn_dump_mount_and_load - swh_vault_cooking - swh_vault_batch_cooking - swh_loader_pypi task_soft_time_limit: 0
After a git bisect section, I have identified the commit in swh-scheduler that breaks the tasks execution: rDSCH3c41cef4a83b37d3174633c23e72a52d18d1ec70