Skip to content

Fix GCP Batch co-execution deadlocks #466

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mvdbeek merged 4 commits into from
Jul 3, 2026
Merged

Fix GCP Batch co-execution deadlocks #466

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
1963eb3
Enable background task
ksuderman Feb 16, 2026
3c90167
Pass docker_extra_volumes to GCP Batch container bind mounts
ksuderman Mar 4, 2026
e194624
Make tool container background and sidecar foreground in GCP Batch
ksuderman Mar 6, 2026
bcd073e
Fix GCP Batch runnable ordering: tool (bg) first, sidecar (fg) second
ksuderman Mar 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 22 additions & 2 deletions pulsar/client/client.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1120,13 +1120,33 @@ def _launch_containers(
assert pulsar_finish_container is None
gcp_job_params = self._gcp_job_params
job = gcp_job_template(gcp_job_params)
runnable = container_command_to_gcp_runnable("pulsar-container", pulsar_submit_container)
job.task_groups[0].task_spec.runnables.append(runnable)

# Parse docker_extra_volumes (comma-separated Docker -v style strings)
# into a list for GCP Batch Runnable.Container.volumes
extra_volumes = []
raw = self.destination_params.get("docker_extra_volumes", "")
if raw:
extra_volumes = [v.strip() for v in raw.split(",") if v.strip()]

# Order matters: GCP Batch runs runnables sequentially. A background
# runnable starts and immediately yields to the next runnable. We need:
# 1. Tool (background) — starts polling for command_line file
# 2. Sidecar (foreground) — stages inputs, writes command_line, polls
# for return_code, collects outputs, sends AMQP callback
# The sidecar must be foreground so it survives after the tool finishes
# (GCP Batch kills background runnables when all foreground ones exit).
if tool_container:
tool_runnable = container_command_to_gcp_runnable("tool-container", tool_container)
tool_runnable.background = True
if extra_volumes:
tool_runnable.container.volumes = extra_volumes
job.task_groups[0].task_spec.runnables.append(tool_runnable)

runnable = container_command_to_gcp_runnable("pulsar-container", pulsar_submit_container)
if extra_volumes:
runnable.container.volumes = extra_volumes
job.task_groups[0].task_spec.runnables.append(runnable)

job_name = self._job_name
create_request = gcp_job_request(gcp_job_params, job, job_name)
client = gcp_client(gcp_job_params.credentials_file)
Expand Down