jcreek/CosmicClash
1
0
Fork 0
mirror of https://github.com/jcreek/CosmicClash.git synced 2026-09-10 16:04:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(multiplayer): recover recorded allocations

Browse Source
This commit is contained in:
Josh Creek
2026-09-01 10:40:10 +01:00
parent 55c46f56ec
commit 25fdc2c2c8
6 changed files with 68 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
multiplayer-next.md
+4 -3
View File
@@ -67,7 +67,8 @@ product policy are in [`docs/MATCHMAKING.md`](docs/MATCHMAKING.md).
publication now requires an allocated endpoint and verifies canonical
join-authorisation signatures before exposing player rows; `cmd/allocator`
now polls these claims and drives provider allocation/reconciliation/binding;
interrupted-provider reconciliation and live Agones integration remain.
durable post-provider recovery avoids a second allocation after a bind crash;
unknown provider-outcome reconciliation and live Agones integration remain.
- [ ] **IN PROGRESS:** Run the Go control plane against PostgreSQL/Redis with
independently runnable API, matcher, allocator and maintenance roles. The
`cmd/control-plane` API role now opens PostgreSQL, applies migrations, wires
@@ -79,8 +80,8 @@ product policy are in [`docs/MATCHMAKING.md`](docs/MATCHMAKING.md).
allocating matches and invokes Agones through the durable allocator boundary;
`cmd/maintenance` now runs
bounded ranked-season rollover batches with signal-bound shutdown;
interrupted-provider reconciliation, Redis failover and live service checks
remain.
unknown provider-outcome reconciliation, Redis failover and live service
checks remain.
- [ ] **IN PROGRESS:** Define assignment compatibility and opt-in `ServerConfig`
flags whose defaults reproduce the community-server path. Allocation manifest
validation now covers client build and future expiry; allocated servers now
multiplayer-todo.md
+1 -1
View File
@@ -1213,7 +1213,7 @@ the local/CI/community transport, not a silent production fallback.
| 8.27 `[D:8.26]` | **IN PROGRESS.** Go supervisor package provides local-safe Agones REST discovery, validates assigned address/port data, injects dynamic `SDR_LISTEN_PORT`/`SDR_IP`, performs explicit process-ready probing and Ready transition; direct mode bypasses Agones | `server/supervisor/` covers allocated/direct startup, invalid endpoint rejection, dynamic endpoint/Ready ordering and authenticated drain; allocated Godot now supplies a loopback readiness/drain control surface and `agones_sdk.gd` supplies sidecar Health/Ready/Shutdown/annotation REST operations; metadata watch, real Agones annotation/shutdown confirmation and emulator integration remain |
| 8.28 `[D:8.6,8.27]` | **IN PROGRESS.** Supervisor separates explicit process-ready from Agones Ready and never scrapes stdout; allocated mode refuses to mark Ready without a configured readiness probe | `server/supervisor/` tests prove Ready follows the probe and direct mode remains functional; `server_control.gd`, `agones_sdk.gd` and process-level smokes prove loopback `/ready`, `/health`, bearer-protected `/drain`, sidecar-shaped Health/Ready calls and drain admission fencing; detached-container and Health-reclaim integration remain |
| 8.29 `[D:8.26,8.27]` | **IN PROGRESS.** Supervisor discovers and validates the Agones endpoint, propagates the actual dynamic `--port`, and exports `SDR_LISTEN_PORT`/`SDR_IP` only for Hosted-SDR while preserving an isolated ENet path | `server/supervisor/` tests cover invalid address/port rejection, dynamic port argument/env propagation and SDR-vs-ENet separation; real Agones dynamic/passthrough mapping, POP/cert/firewall/NAT and multi-match fixture remain |
| 8.30 `[D:8.18,8.26,8.28,8.29]` | **IN PROGRESS.** Pure Go allocator filters Ready servers by region/build/protocol/transport, atomically claims one with idempotent allocation replay, verifies unanimous accepted-proposal/playlist/participant invariants before provider invocation, and now owns the assignment-publication boundary; PostgreSQL adds durable GameServer registration and compatible `SKIP LOCKED` claims with request-digest fencing, plus a leased `ALLOCATING`-match claim that derives the immutable compatibility tuple and fences bind/release by deterministic allocation ID; `server/agones` submits and validates namespaced `GameServerAllocation` responses and dynamic endpoints; `server/allocator` reconciles provider success into durable state before exposing the endpoint; `cmd/allocator` is a signal-bound polling role that drives the lease → provider → durable-record → match-bind sequence | `server/domain/allocator.go`, `server/store/allocator_sql.go`, `server/store/allocation_match_sql.go`, `server/store/allocation_match_adapter.go`, `server/agones/allocation.go`, `server/allocator/service.go`, `server/allocator/worker.go`, `server/cmd/allocator`, `server/migrations/0004_allocator_registry.sql`, `0006_match_allocation_claims.sql` and tests cover deterministic compatible selection, exhaustion, conflicting/identical allocation replay, allocation-match lease recovery/bind fencing, immutable Fleet selector labels, provider ambiguity lease retention, unknown allocations, SQL claim ordering, invalid input, provider error/malformed response/IPv6 endpoint handling, accepted-proposal gating, durable-reconciliation failure isolation and assignment replay/conflict; `TestPostgreSQLAllocatorClaimReplayAndCapacityFence` and `TestPostgreSQLAllocationMatchClaimLeaseAndBindFence` cover the live database paths when the disposable database gate is run; interrupted-provider reconciliation, signed roster metadata, bounded cross-replica retry and live Agones integration remain |
| 8.30 `[D:8.18,8.26,8.28,8.29]` | **IN PROGRESS.** Pure Go allocator filters Ready servers by region/build/protocol/transport, atomically claims one with idempotent allocation replay, verifies unanimous accepted-proposal/playlist/participant invariants before provider invocation, and now owns the assignment-publication boundary; PostgreSQL adds durable GameServer registration and compatible `SKIP LOCKED` claims with request-digest fencing, plus a leased `ALLOCATING`-match claim that derives the immutable compatibility tuple and fences bind/release by deterministic allocation ID; `server/agones` submits and validates namespaced `GameServerAllocation` responses and dynamic endpoints; `server/allocator` reconciles provider success into durable state before exposing the endpoint; `cmd/allocator` is a signal-bound polling role that drives the lease → provider → durable-record → match-bind sequence and rebinds a recovered durable provider allocation without a second provider call | `server/domain/allocator.go`, `server/store/allocator_sql.go`, `server/store/allocation_match_sql.go`, `server/store/allocation_match_adapter.go`, `server/agones/allocation.go`, `server/allocator/service.go`, `server/allocator/worker.go`, `server/cmd/allocator`, `server/migrations/0004_allocator_registry.sql`, `0006_match_allocation_claims.sql` and tests cover deterministic compatible selection, exhaustion, conflicting/identical allocation replay, allocation-match lease recovery/bind fencing, recovered-allocation binding without provider recall, immutable Fleet selector labels, provider ambiguity lease retention, unknown allocations, SQL claim ordering, invalid input, provider error/malformed response/IPv6 endpoint handling, accepted-proposal gating, durable-reconciliation failure isolation and assignment replay/conflict; `TestPostgreSQLAllocatorClaimReplayAndCapacityFence` and `TestPostgreSQLAllocationMatchClaimLeaseAndBindFence` cover the live database paths when the disposable database gate is run; unknown provider-outcome reconciliation, signed roster metadata, bounded cross-replica retry and live Agones integration remain |
| 8.31 `[D:8.9,8.30]` | **IN PROGRESS.** Pure Go assignment gate requires Allocated state, exact allocation ID/match/server/region/build/protocol/transport compatibility, non-empty hosted endpoint and verified manifest signature before exposure; allocator publication cannot expose Ready state; durable roster persistence now verifies each canonical join-authorisation signature before publishing player rows; allocator service gates roster publication on allocated state and endpoint presence | `server/domain/assignment.go`, `allocator.go`, `store/assignment_sql.go` plus adversarial fixtures cover early-connect, tampered signature/manifest, wrong compatibility, empty endpoint, unknown allocation, forged roster signature, valid signature, premature publication and post-publication mutation rejection; Agones metadata watch, hosted-address registration, production signer and client-ticket publication remain |
| 8.32 `[D:8.2,8.26,8.30]` | **IN PROGRESS.** Provider-neutral FleetAutoscaler baseline preserves a two-process Ready buffer, caps warm capacity, and leaves Allocated scale-down independent of the Ready floor; Fleet image references remain digest-pinned for current/rollback pre-pull | `deploy/k8s/base/fleet-autoscaler.yaml` and manifest tests cover Fleet ownership, Buffer policy and floor/cap invariants; regional on-demand node pools/failure domains, pre-pull rollout, warm-allocation p95/p99 and N+1 certification remain |
| 8.33 `[D:8.26,8.32]` | **IN PROGRESS.** Fleet scheduling now requires on-demand capacity and spreads Ready processes across zones with skew 1; the autoscaler preserves the two-process Ready floor | `deploy/k8s/base/fleet.yaml` and manifest tests reject interruptible placement and single-zone concentration structurally; regional node pools, forced node-loss testing and measured N+1 headroom remain |
server/allocator/worker.go
+11 -3
View File
@@ -13,6 +13,7 @@ import (
// lease a match before returning it and fence binding by allocation ID.
type MatchClaimSource interface {
ClaimAllocatingMatch(context.Context, time.Time) (domain.AllocationRequest, bool, error)
FindProviderAllocation(context.Context, domain.AllocationRequest) (domain.Allocation, bool, error)
BindAllocatedMatch(context.Context, domain.Allocation) error
}
@@ -37,11 +38,18 @@ func (w Worker) RunOnce(ctx context.Context) (bool, error) {
if err != nil || !found {
return found, err
}
result, err := w.Service.Allocate(ctx, request, AllocationLabels(request))
allocation, recorded, err := w.Claims.FindProviderAllocation(ctx, request)
if err != nil {
return true, fmt.Errorf("allocate claimed match %s: %w", request.MatchID, err)
return true, fmt.Errorf("recover allocation for match %s: %w", request.MatchID, err)
}
if err := w.Claims.BindAllocatedMatch(ctx, result.Allocation); err != nil {
if !recorded {
result, err := w.Service.Allocate(ctx, request, AllocationLabels(request))
if err != nil {
return true, fmt.Errorf("allocate claimed match %s: %w", request.MatchID, err)
}
allocation = result.Allocation
}
if err := w.Claims.BindAllocatedMatch(ctx, allocation); err != nil {
return true, fmt.Errorf("bind allocated match %s: %w", request.MatchID, err)
}
return true, nil
server/allocator/worker_test.go
+23 -5
View File
@@ -12,11 +12,17 @@ import (
)
type matchClaimSpy struct {
request domain.AllocationRequest
found bool
err error
bound domain.Allocation
bindErr error
request domain.AllocationRequest
found bool
err error
recorded domain.Allocation
recordErr error
bound domain.Allocation
bindErr error
}
func (s *matchClaimSpy) FindProviderAllocation(_ context.Context, _ domain.AllocationRequest) (domain.Allocation, bool, error) {
return s.recorded, s.recorded.AllocationID != "", s.recordErr
}
func (s *matchClaimSpy) ClaimAllocatingMatch(_ context.Context, _ time.Time) (domain.AllocationRequest, bool, error) {
@@ -51,6 +57,18 @@ func TestWorkerRetainsClaimWhenProviderOutcomeIsAmbiguous(t *testing.T) {
}
}
func TestWorkerRecoversDurableProviderAllocationWithoutCallingProvider(t *testing.T) {
request := domain.AllocationRequest{AllocationID: "allocation-1", MatchID: "match-1", Region: "EU", Build: "build-1", Protocol: 1, Transport: "enet"}
recorded := domain.Allocation{AllocationID: request.AllocationID, MatchID: request.MatchID, ServerID: "server-1", Region: request.Region, Build: request.Build, Protocol: request.Protocol, Transport: request.Transport, State: domain.ServerAllocated}
claims := &matchClaimSpy{request: request, found: true, recorded: recorded}
provider := &providerSpy{}
worker := Worker{Claims: claims, Service: Service{Provider: provider, Durable: &durableSpy{}, Now: func() time.Time { return time.Unix(1_000, 0) }}, Now: func() time.Time { return time.Unix(1_000, 0) }}
processed, err := worker.RunOnce(context.Background())
if err != nil || !processed || provider.calls != 0 || claims.bound != recorded {
t.Fatalf("processed=%t err=%v provider=%d bound=%+v", processed, err, provider.calls, claims.bound)
}
}
func TestWorkerDoesNothingWhenNoDurableMatchIsAvailable(t *testing.T) {
claims := &matchClaimSpy{}
worker := Worker{Claims: claims, Now: func() time.Time { return time.Unix(1_000, 0) }}
server/store/allocation_match_adapter.go
+4
View File
@@ -20,6 +20,10 @@ func (s AllocatingMatchClaims) ClaimAllocatingMatch(ctx context.Context, now tim
return item.Request, found, err
}
func (s AllocatingMatchClaims) FindProviderAllocation(ctx context.Context, request domain.AllocationRequest) (domain.Allocation, bool, error) {
return FindProviderAllocation(ctx, s.DB, request)
}
func (s AllocatingMatchClaims) BindAllocatedMatch(ctx context.Context, allocation domain.Allocation) error {
return BindAllocatedMatch(ctx, s.DB, allocation)
}
server/store/allocation_match_sql.go
+25
View File
@@ -1,6 +1,7 @@
package store
import (
"bytes"
"context"
"database/sql"
"fmt"
@@ -47,6 +48,30 @@ const ReleaseAllocatedMatchClaimSQL = `UPDATE matches
SET allocation_id = NULL, allocation_claimed_at = NULL
WHERE match_id = $1 AND state = 'ALLOCATING' AND allocation_id = $2 AND server_id IS NULL`
// FindProviderAllocation verifies whether a recovered lease has already
// crossed the durable provider boundary. A worker can then bind it without
// issuing a second external allocation request after a crash.
func FindProviderAllocation(ctx context.Context, db *sql.DB, request domain.AllocationRequest) (domain.Allocation, bool, error) {
if db == nil || request.AllocationID == "" || request.MatchID == "" || request.Region == "" || request.Build == "" || request.Protocol <= 0 || (request.Transport != "enet" && request.Transport != "steam_sdr") {
return domain.Allocation{}, false, fmt.Errorf("invalid provider allocation lookup")
}
var allocation domain.Allocation
var digest []byte
err := db.QueryRowContext(ctx, SelectAllocationSQL, request.AllocationID).Scan(&allocation.AllocationID, &allocation.MatchID, &allocation.ServerID, &allocation.Region, &allocation.Build, &allocation.Protocol, &allocation.Transport, &allocation.AllocatedAt, &digest)
if err == sql.ErrNoRows {
return domain.Allocation{}, false, nil
}
if err != nil {
return domain.Allocation{}, false, err
}
want := allocationRequestDigest(request)
if !bytes.Equal(digest, want[:]) || allocation.MatchID != request.MatchID || allocation.Region != request.Region || allocation.Build != request.Build || allocation.Protocol != request.Protocol || allocation.Transport != request.Transport {
return domain.Allocation{}, false, domain.ErrConflict
}
allocation.State = domain.ServerAllocated
return allocation, true, nil
}
// ClaimAllocatingMatch returns one durable provider work item. The fixed
// allocation ID is retained across a lease recovery, allowing every later
// reconciliation step to reject a different server for the same match.