feat: implement health-aware restart logic for remote workloads #3433
+191
−46
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit enhances the restart logic for remote workloads (like Glean/Datadog) to be health-aware. Instead of just checking if the supervisor process exists (PID check), it now also performs an HTTP health check against the proxy. This addresses the "alive but unresponsive" state that often occurs after laptop sleep/wake cycles, where the process exists but the network socket is broken or the application is hung. Changes: - Added field to for dependency injection - Updated to perform HTTP health checks for remote workloads - Updated tests to mock health checks and verify correct behavior - Fixed test expectations for workload name vs base name usage in PID checks This is a step towards a more robust supervisor that can auto-heal broken workloads. Signed-off-by: Frederic Le Feurmou <flfeurmou@indeed.com>
dmjb
requested changes
Jan 27, 2026
Member
dmjb
left a comment
dmjb
left a comment
There was a problem hiding this comment.
We have healthchecks implemented elsewhere in the codebase, but we removed them to fix another bug. We have an issue open to reinstate them, I would prefer to close the PR and fix the original healthcheck mechanism.
| @@ -809,25 +775,125 @@ func (d *DefaultManager) getWorkloadContainer(ctx context.Context, name string) | |||
| return &container, nil | |||
| } | |||
|
|
|||
| // checkHTTPHealth performs a simple HTTP GET request to check health | |||
| func checkHTTPHealth(ctx context.Context, url string) bool { | |||
Member
There was a problem hiding this comment.
We have these sorts of health checks implemented further down the stack (in the proxy and transport layer).
Contributor
Author
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This PR implements health-aware restart logic for remote workloads (like Glean and Datadog). It enhances the supervisor check to verify not just that the process exists, but that it is responsive to HTTP health checks.
Problem
After laptop sleep/wake cycles, remote MCP server proxies often end up in a "zombie" state where the process exists (PID is valid) but the network socket is broken or the application is hung. The current logic only checks for PID existence, so it assumes the workload is healthy and refuses to restart it, or fails to clean it up properly.
Solution
Updated
isSupervisorHealthy(formerlyisSupervisorProcessAlive) to:process.FindProcess)falseif either check fails, triggering the restart/cleanup logicThis allows ToolHive to automatically detect and recover from "alive but unresponsive" states.
Changes
DefaultManagerto support dependency injection for health checkscheckHTTPHealthhelperisSupervisorHealthyto use the new checkTesting