feat(BA-6894): implement network timeout idle checker#12878
Open
seedspirit wants to merge 3 commits into
Open
Conversation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
github-actions
Bot
added
comp:manager
…ons_batch Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Adds reconciler-based network timeout checking for interactive sessions using Valkey activity signals.
Changes:
- Implements configurable network-idle judgments.
- Adds batched active-connection counting.
- Wires Valkey dependencies and adds tests.
Reviewed changes
Copilot reviewed 17 out of 17 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
tests/unit/manager/sokovan/idle_check/test_stage.py |
Updates stage dependency tests. |
tests/unit/manager/sokovan/idle_check/test_session_lifetime_checker.py |
Updates network spec fixture. |
tests/unit/manager/sokovan/idle_check/test_network_timeout_checker.py |
Tests timeout judgments and validation. |
tests/unit/manager/sokovan/idle_check/test_handler.py |
Updates handler fixtures. |
tests/unit/manager/repositories/idle_checker/test_repository.py |
Updates persisted test specification. |
tests/unit/manager/dependencies/orchestration/test_sokovan.py |
Supplies Valkey live dependency. |
tests/unit/manager/dependencies/orchestration/test_composer.py |
Updates orchestration composition tests. |
tests/unit/common/clients/valkey_client/test_valkey_live_client.py |
Tests batched connection counts. |
src/ai/backend/manager/sokovan/stages/idle_check.py |
Registers the network checker. |
src/ai/backend/manager/sokovan/stages/factory.py |
Passes Valkey into the stage. |
src/ai/backend/manager/sokovan/idle_check/checkers/network_timeout.py |
Implements network-idle evaluation. |
src/ai/backend/manager/dependencies/orchestration/sokovan.py |
Adds orchestrator Valkey input. |
src/ai/backend/manager/dependencies/orchestration/composer.py |
Propagates Valkey through orchestration. |
src/ai/backend/manager/dependencies/composer.py |
Supplies infrastructure Valkey client. |
src/ai/backend/common/data/idle_checker/types.py |
Defines network timeout configuration. |
src/ai/backend/common/clients/valkey_client/valkey_live/client.py |
Adds batched connection counting. |
changes/12878.feature.md |
Adds the feature changelog entry. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+99
to
+101
| self._valkey_live.count_active_connections_batch([ | ||
| str(session_id) for session_id in session_ids | ||
| ]), |
Comment on lines
+69
to
+70
| if state.last_access is None: | ||
| continue |
Comment on lines
+46
to
+66
| # Fetch session states in one batch to avoid repeated I/O per assignment. | ||
| sessions: list[IdleCheckSession] = [] | ||
| for assignment in assignments: | ||
| sessions.extend(assignment.sessions) | ||
| states = await self._prepare_states(sessions) | ||
|
|
||
| # Judge each assignment in one pass, using the pre-fetched states. | ||
| judgments: list[IdleJudgment] = [] | ||
| for assignment in assignments: | ||
| network_spec = assignment.definition.spec.network | ||
| if network_spec is None: | ||
| log.error( | ||
| "Network timeout checker {} has mismatched spec type: {}", | ||
| assignment.definition.checker_id, | ||
| assignment.definition.spec.type, | ||
| ) | ||
| continue | ||
| # Skip sessions with no timeout configured (0 means "never idle"). | ||
| if network_spec.idle_timeout_seconds == 0: | ||
| continue | ||
| idle_timeout_seconds = Decimal(network_spec.idle_timeout_seconds) |
fregataa
reviewed
Jul 15, 2026
| if state.last_access is None: | ||
| continue | ||
| current_time = Decimal(str(context.current_time.timestamp())) | ||
| idle_seconds = (current_time - state.last_access).normalize() |
Member
There was a problem hiding this comment.
Are current_time and last_access from the same source?
Contributor
Author
There was a problem hiding this comment.
"current time" is injected by the idle checker, and "last_access" is the value stored in valkey.
Comment on lines
+95
to
+100
| last_access_values, active_connection_counts = await asyncio.gather( | ||
| self._valkey_live.get_multiple_live_data([ | ||
| f"session.{session_id}.last_access" for session_id in session_ids | ||
| ]), | ||
| self._valkey_live.count_active_connections_batch(session_ids), | ||
| ) |
Member
There was a problem hiding this comment.
Are they grouped for concurrent query?
Contributor
Author
There was a problem hiding this comment.
Could have run the queries separately, but I used gather because we thought running them simultaneously would yield slightly higher accuracy. It’s not required, so we can run them separately if you prefer.
This was referenced Jul 15, 2026
Closed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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
NetworkTimeoutCheckerto the sokovan reconciler idle-check stage: judges interactive sessions idle when they have no active app connections and the last network access exceeds the configured timeout.NetworkTimeoutSpec.idle_timeout_seconds(0 disables the checker definition) and register the checker via DI (valkey_livewired through orchestration composers tobuild_idle_check_stage).ValkeyLiveClient.count_active_connections_batchto fetch per-session connection counts in a single batch alongside batched last-access reads.Test plan
pants testontest_network_timeout_checker.py,test_stage.py,test_valkey_live_client.py(checker judgments, DI wiring, batch count against real Valkey)pants fmt/lint/checkon changed filesResolves BA-6894
🤖 Generated with Claude Code