fix(acms): harden hot/warm/cold tier service reliability #11238
Merged
HAL9000
merged 2 commits from 2026-05-28 21:31:23 +00:00
pr-fix/9663-hot-warm-cold-tier-reliability into master
Labels
Clear labels
auto/needs-reevaluation
controller-managed
overdue
auto/blocked-by-deps
auto/ci-timeout
auto/claimed-implementer
auto/claimed-merge
auto/claimed-reviewer
auto/driver-down
auto/invariant-violation
auto/last-attempt-tier-0
auto/last-attempt-tier-1
auto/last-attempt-tier-2
auto/last-attempt-tier-min
Automation Tracking
auto/needs-conflict-resolution
auto/needs-implementer
auto/postmortem
auto/ready-to-merge
auto/restart-throttled
auto/revert
auto/sentinel
auto/stale-inactivity
auto/unstable
Blocked
Needs Feedback
Signed-off: Owner
Signed-off: Scrum Master
Signed-off: Tech Lead
Spike
Controller deferred this PR; awaiting Phase 6+ scope-evaluator or operator re-enablement.
Auto-agents controller manages this PR/issue (see tools/controller/deploy/RUNBOOK.md). Remove this label to abandon controller management.
PR blocked by an open issue dependency. Operator must close the dep (or remove the dependency link) before the merge driver can act. Auto-cleared by merge_drive when no open deps remain.
Most recent merge cycle hit CI timeout. Driver excludes this PR while last merge_cycle row is < 30 min old; label persists thereafter as visible history.
Currently being processed by an implementer worker.
Currently being processed by the merge driver.
Currently being processed by a reviewer worker.
Merge driver heartbeat stale; pipeline halted. Closed automatically on next clean tick.
Detected master commit violating the strict merge invariant. Tracked as an issue (not a PR label); kept here for label completeness.
In-cycle escalation: most recent attempt ran at the Tier 0 slot (`tier-0`). Slot's model defined in .opencode/models/tiers.yaml.
In-cycle escalation: most recent attempt ran at the Tier 1 slot (`tier-1`). Slot's model defined in .opencode/models/tiers.yaml.
In-cycle escalation: most recent attempt ran at the Tier 2 slot (`tier-2`). Slot's model defined in .opencode/models/tiers.yaml. Gated behind IMPLEMENTER_ESCALATION_TIER2_ENABLED.
In-cycle escalation: most recent attempt ran at the Tier -1 slot (`tier-min`). Slot's model defined in .opencode/models/tiers.yaml. Suffix is ``-min`` (not ``--1``) so the Forgejo UI reads naturally.
Tracking issues used by the AI Automation system for agents to communicate and report.
Rebase conflict needs LLM conflict-resolver.
Failing CI needs implementer attention.
Documenting a driver incident or rollback.
Reviewer has APPROVED this PR and no later REQUEST_CHANGES is outstanding. The merge driver requires this label to even consider a PR for merging. Set by the reviewer worker on APPROVE; cleared on REQUEST_CHANGES.
Train repeatedly lost master-tempo races. Driver excludes via merge_cycle until cooldown elapses; label persists as visible history.
Revert PR backing out an invariant violation. Fast-tracked through the merge driver.
Sentinel PR duplicated from upstream into a personal fork by tools/duplicate_prs_to_fork.py for pipeline testing. Lives only in the fork; the canonical pipeline never sees it.
No implementer activity for N days. Flagged for human review. Auto-cleared on next push to head branch.
Repeatedly fails on current master (>= 3 ci-fail-on-rebased-sha releases in 12 h). Excluded from driver until human triage.
A ticket in a blocked state and unable to complete until some other task is completed first.
Bounty
$100
A bounty of $100 for any open-source contributor who provides a MR that solves this issue
Bounty
$1000
A bounty of $1000 for any open-source contributor who provides a MR that solves this issue
Bounty
$10000
A bounty of $10000 for any open-source contributor who provides a MR that solves this issue
Bounty
$20
A bounty of $20 for any open-source contributor who provides a MR that solves this issue
Bounty
$2000
A bounty of $2000 for any open-source contributor who provides a MR that solves this issue
Bounty
$250
A bounty of $250 for any open-source contributor who provides a MR that solves this issue
Bounty
$50
A bounty of $50 for any open-source contributor who provides a MR that solves this issue
Bounty
$500
A bounty of $500 for any open-source contributor who provides a MR that solves this issue
Bounty
$5000
A bounty of $5000 for any open-source contributor who provides a MR that solves this issue
Bounty
$750
A bounty of $750 for any open-source contributor who provides a MR that solves this issue
MoSCoW
Could have
Could have feature in order to satisfy the epic/legendary.
MoSCoW
Must have
Must have feature in order to satisfy the epic/legendary.
MoSCoW
Should have
Should have feature in order to satisfy the epic/legendary.
There are questions in the ticket that can not be completed until the project owner provides clarity.
Points
1
1 man-hours worth of work for an expert with no learning curve.
Points
13
13 man-hours worth of work for an expert with no learning curve.
Points
2
2 man-hours worth of work for an expert with no learning curve.
Points
21
21 man-hours worth of work for an expert with no learning curve.
Points
3
3 man-hours worth of work for an expert with no learning curve.
Points
34
34 man-hours worth of work for an expert with no learning curve.
Points
5
5 man-hours worth of work for an expert with no learning curve.
Points
55
55 man-hours worth of work for an expert with no learning curve.
Points
8
8 man-hours worth of work for an expert with no learning curve.
Points
88
88 man-hours worth of work for an expert with no learning curve.
Priority
Backlog
This ticket has backlogged priority and is not to be worked on yet
Priority
CI Blocker
Critical priority issue that blocks CI/CD pipeline and prevents PR merges
Priority
Critical
The priority is critical
Priority
High
The priority is high
Priority
Low
The priority is low
Priority
Medium
The priority is medium
When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.
When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.
When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.
A ticket for learning a tool or technology that is needed to be able to do future planning and design.
State
Completed
The ticket has been fully implemented, completed, and merged with the source code. This label should only be applied once a ticket is closed.
State
Duplicate
A ticket that represents the same content as an existing ticket.
State
In Progress
A ticket that is actively being developed.
State
In Review
A ticket that has had some code completed to implement but is waiting to pass peer review and is not yet merged in.
State
Paused
This ticket's work started but wasn't finished. It's on hold (likely in a feature branch) and will be resumed later, either due to a blocker or a delay.
State
Unverified
All new tickets start in this state. A developer may set it to show the ticket is unverified. This means we haven't agreed to work on it. It will either move to a verified state or be closed as wontdo.
State
Verified
The issue has been verified by a developer as legitimate. It will be worked on and verified tickets are now considered part of the backlog.
State
Wont Do
This ticket has been decided it wont be done. This may mean the bug has been determined to not be real (cant verify) or the feature is one we have decided we dont want to adopt.
Type
Automation
Any edits or discussion about the AI automated coding system.
Type
Bug
Something that doesnt work as intended.
Type
Discussion
Anytime a ticket represents a discussion about a subject and doesnt fall into one of the other categories.
Type
Documentation
An error or improvement needed in the documentation.
Type
Epic
Any first tier epic. That is, an epic which contains only issues as children and will not have sub-epics.
Type
Feature
Some new functionality not present.
Type
Legendary
A type of Epic which will contain other Epics.
Type
Refactor
A code change that restructures existing code without changing its external behavior.
Type
Support
Someone needs help using the project.
Type
Task
A generic task that doesnt fit into the other type categories.
Type
Testing
Work exclusively focusing on fixing or expanding testing.
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
aditya (Aditya Chhabra)
aleenaumair (Aleena Umair)
brent.edwards (Brent Edwards)
CoreRasurae (Luis Mendes)
drew (Drew Morris)
eugen.thaci (Eugen Thaci)
freemo (Jeffrey Phillips Freeman)
HAL9000 (HAL 9000)
HAL9001 (HAL9001)
hamza.khyari (Hamza Khyari)
hurui200320 (Rui Hu)
justin.morris
khird (Kyle Hird)
org.cleveragents
Clear assignees
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: cleveragents/cleveragents-core#11238
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Delete Branch "pr-fix/9663-hot-warm-cold-tier-reliability"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Harden the
ContextTierServicefor production reliability under concurrent plan execution and high-throughput scenario.Changes
Remove dead conflicting defaults: Deleted
_DEFAULT_MAX_TOKENS_HOT,_DEFAULT_MAX_DECISIONS_WARM,_DEFAULT_MAX_DECISIONS_COLDwhich were never used (all budget flows throughbudget_from_settings()incontext_tier_settings.py) and contradicted the canonical values.Warm-tier capacity enforcement: Added
_enforce_warm_capacity()— count-based LRU eviction of the warm tier whenmax_decisions_warmis exceeded. Triggered after:promote())Immutable snapshot returns:
get_all_fragments()andget_hot_fragments()now returnmodel_copy()instances instead of live mutable references, preventing state corruption from external mutation while the RLock is held.Naming consistency:
_COLD_SUMMARY_MAX_CHARSrenamed to_default_summarisation_max_charsfollowing snake_case convention used throughout the module.Impact
TieredFragment).max_decisions_warmcapacity after promotions.03e34b035eto5031170a715031170a71to68fa0479c0Claimed by
merge_drive.py(pid 1264876) until2026-05-28T23:01:17.275801+00:00.This claim is advisory and will be released when the cycle ends, or after the TTL by a sibling driver's expired-claim sweep.
Approved by the controller reviewer stage (workflow 17).
event occurred 2026-05-28T13:54:29.105366+00:00
🌱 Grooming: proceed — PR cleared for processing.
(check
no_duplicates, categoryno_duplicates)PR #11238 is a targeted hardening fix for ContextTierService that combines four distinct reliability improvements: removing dead defaults, enforcing warm-tier capacity with LRU eviction, returning immutable fragment snapshots, and naming consistency. No open PR duplicates this specific combination. Related ACMS PRs (#9663, #10783, #11096) address original feature implementations or different constraints (budget enforcement, path matching, pipeline wiring); none implement the warm-capacity enforcement or immutability patterns central to #11238.
event occurred 2026-05-28T13:57:04.956938+00:00
📋 Estimate: tier 1.
Single-file change (+41/-18) but involves non-trivial new logic: a new LRU eviction method (_enforce_warm_capacity()), changed return semantics for two public methods (mutable refs → model_copy() immutable snapshots), and modifications to the promote() call chain. Thread-safety reasoning required to verify correctness of RLock usage around the new eviction path. Not mechanical — standard engineering work warrants tier 1.
(attempt #3, tier 1)
event occurred 2026-05-28T13:58:30.893624+00:00
🔧 Implementer attempt —
rebased.Pushed 1 commit:
5031170.event occurred 2026-05-28T19:46:37.111234+00:00
🌱 Grooming: proceed — PR cleared for processing.
(check
no_duplicates, categoryno_duplicates)Searched open PRs for topical match on
ContextTierServicehardening. Found related feature PR #9663 (hot/warm/cold tier implementation) but it is the original feature, not a duplicate of this fix. Other ACMS PRs address path matching, pipeline wiring, budget enforcement, and indexing—none overlap with ContextTierService production-reliability hardening. No duplicate detected.event occurred 2026-05-28T19:50:33.790742+00:00
📋 Estimate: tier 1.
Single-file change (+41/-18) to ContextTierService, but contains non-trivial logic: a new _enforce_warm_capacity() method with count-based LRU eviction logic requiring correctness reasoning under concurrent access (RLock), and immutable snapshot returns via model_copy() with thread-safety implications. Dead code removal and constant rename are mechanical, but the eviction and snapshot components involve concurrency semantics that exceed the reliable envelope of tier 0. CI is fully green. Tier 1 is the appropriate default for this class of focused-but-logic-heavy single-file work.
(attempt #7, tier 1)
event occurred 2026-05-28T19:51:15.657461+00:00
🔧 Implementer attempt —
rebased.Pushed 1 commit:
68fa047.event occurred 2026-05-28T20:15:50.347667+00:00
🔴 Changes requested
Confidence: high.
Blocking issues (1):
[error]
src/cleveragents/application/services/context_tiers.py:295-307— Lines 295-307 (cold→warm branch ofpromote):frag = self._cold.pop(fragment_id)
promoted = frag.model_copy(update={"tier": ContextTier.WARM})
self._warm[fragment_id] = promoted
self._emit_tier_event(...)
Enforce warm-tier capacity after promotion.
self._enforce_warm_capacity()
return promoted
After
_enforce_warm_capacity()runs it callsdel self._warm[oldest_id]on the LRU warm fragment (lines 461). When the warm tier is at capacity and the just-promoted cold fragment has the oldestlast_accessedvalue among all warm fragments (no_touch()is called during cold→warm promotion — compare to_touchin the_touchstaticmethod at line 518), the newly added fragment is the eviction target. It was already removed fromself._coldat line 296, so after the eviction it exists in no tier. Yet line 307 unconditionally returnspromotedwith.tier=ContextTier.WARM, giving the caller a stale handle to a fragment that is silently lost from all stores. The warm→hot path at lines 322-335 explicitly guards against this identical scenario withif fragment_id not in self._hot:followed by restoration +_enforce_warm_capacity(). The cold→warm path was not given the same protection.self._enforce_warm_capacity()at line 306, add:if fragment_id not in self._warm: promoted = self._touch(promoted); self._warm[fragment_id] = promoted; self._enforce_warm_capacity(). The_touch()call updateslast_accessedto now so the restored fragment is not immediately re-evicted. Alternatively, call_touch()onpromotedbefore inserting at line 298 (i.e.promoted = self._touch(frag.model_copy(update={"tier": ContextTier.WARM}))), ensuring the just-promoted fragment is never the LRU immediately after promotion — matching the implicit access semantics of a promotion event.(attempt #9, tier 1)
event occurred 2026-05-28T20:55:06.953674+00:00
🔧 Implementer attempt —
resolved.Pushed 1 commit:
0fee675.Files touched:
src/cleveragents/application/services/context_tiers.py.event occurred 2026-05-28T21:24:46.362843+00:00
✅ Approved
Reviewed at commit
0fee675.Confidence: high.