Release v0.2.0: complete platform hardening, autoscale/keep-alive robustness, logging overhaul, and drop-in compatibility fixes

Author: Seggys116

.env.example

@@ -22,6 +22,12 @@ URT_PORT=80
 # Fallback: OPR_EXECUTOR_SECRET
 URT_SECRET=your-secret-key-here
 
+# Enable Prometheus metrics endpoint at /metrics
+# Values: true/false, 1/0, yes/no, on/off, enabled/disabled
+# Default: false
+# Fallback: OPR_EXECUTOR_METRICS
+URT_METRICS=false
+
 # =============================================================================
 # RESOURCE OVERRIDES
 # =============================================================================
@@ -64,13 +70,36 @@ URT_INACTIVE_THRESHOLD=60
 # Fallback: OPR_EXECUTOR_MAINTENANCE_INTERVAL
 URT_MAINTENANCE_INTERVAL=3600
 
+# Autoscale mode (recommended for multi-replica deployments)
+# Enables adaptive concurrency guards to smooth burst traffic.
+# Fallback: OPR_EXECUTOR_AUTOSCALE
+URT_AUTOSCALE=false
+
+# Optional hard limit for concurrent executions when autoscale is enabled.
+# Set to 0 or leave empty for adaptive default.
+# Fallback: OPR_EXECUTOR_MAX_CONCURRENT_EXECUTIONS
+URT_MAX_CONCURRENT_EXECUTIONS=
+
+# Optional hard limit for concurrent runtime creations when autoscale is enabled.
+# Set to 0 or leave empty for adaptive default.
+# Fallback: OPR_EXECUTOR_MAX_CONCURRENT_RUNTIME_CREATES
+URT_MAX_CONCURRENT_RUNTIME_CREATES=
+
+# Max milliseconds to wait for an execution queue slot before fast-failing.
+# Fallback: OPR_EXECUTOR_EXECUTION_QUEUE_WAIT_MS
+URT_EXECUTION_QUEUE_WAIT_MS=2000
+
+# Max milliseconds to wait for a runtime-create queue slot before fast-failing.
+# Fallback: OPR_EXECUTOR_RUNTIME_CREATE_QUEUE_WAIT_MS
+URT_RUNTIME_CREATE_QUEUE_WAIT_MS=5000
+
 # =============================================================================
 # DOCKER CONFIGURATION
 # =============================================================================
 
 # Docker network name(s) for runtime containers (comma-separated)
 # Fallback: OPR_EXECUTOR_NETWORK
-URT_NETWORK=executor_runtimes
+URT_NETWORK=openruntimes-runtimes
 
 # Enable/disable auto-pulling images on startup
 # Set to "disabled" to skip image pulls
@@ -79,7 +108,7 @@ URT_IMAGE_PULL=enabled
 
 # Comma-separated list of allowed runtime images
 # Leave empty to allow all images
-# Fallback: OPR_EXECUTOR_RUNTIMES
+# Fallbacks: OPR_EXECUTOR_RUNTIMES, OPR_EXECUTOR_IMAGES
 URT_RUNTIMES=
 
 # Supported runtime versions (comma-separated)
@@ -120,3 +149,7 @@ URT_RETRY_DELAY_MS=500
 
 # Log level: trace, debug, info, warn, error
 RUST_LOG=info
+
+# Optional: clean local storage cache on shutdown.
+# Default false preserves warm-cache artifacts between restarts.
+URT_CACHE_CLEANUP_ON_SHUTDOWN=false

CHANGELOG.md

@@ -2,6 +2,27 @@
 
 All notable changes to this project will be documented in this file.
 
+## [Unreleased]
+
+## [0.2.0] - 2026-02-25
+
+### Changed
+- **Autoscale controls**: Added `URT_AUTOSCALE`, `URT_MAX_CONCURRENT_EXECUTIONS`, and `URT_MAX_CONCURRENT_RUNTIME_CREATES` to enable adaptive runtime/execution concurrency limiting without API schema changes.
+- **Load shedding and queue control**: Added bounded queue wait controls via `URT_EXECUTION_QUEUE_WAIT_MS` and `URT_RUNTIME_CREATE_QUEUE_WAIT_MS`, returning fast overload responses (`429`/`503`) under sustained saturation.
+- **Transient retry hardening**: Added jittered exponential backoff with transient/non-transient error classification for runtime source downloads, runtime container creation, build artifact uploads, and runtime execution protocol calls.
+- **Metrics expansion**: `/metrics` now exports persistent counters/histograms for queue wait/depth, execution/runtime-create latency, retries, keep-alive transfer and cleanup outcomes, active executions, and error classes.
+- **Keep-alive replacement serialization**: Added per-`keepAliveId` async locking to serialize ownership transfer, replacement cleanup, deletion, and maintenance reconciliation paths.
+- **Keep-alive generation labels**: New runtime containers include `urt.keep_alive_generation` and cleanup uses owner/generation guards to avoid removing the active replacement.
+- **Prometheus exporter support**: Added optional `/metrics` endpoint for Prometheus/Grafana, toggled by `URT_METRICS` (with `OPR_EXECUTOR_METRICS` fallback), and protected by bearer auth when `URT_SECRET` is configured.
+- **Request log correlation**: Added structured request logging with propagated/generated `x-request-id` response headers.
+- **Startup/warmup noise reduction**: Runtime/network allowlists are deduplicated and startup network attach now skips unresolved executor container names instead of emitting repeated missing-container warnings.
+- **Cache lifecycle behavior**: Added `URT_CACHE_CLEANUP_ON_SHUTDOWN` (default `false`) so warm cache can persist across restarts for lower cold-start latency.
+- **Drop-in startup compatibility**: `URT_NETWORK` now defaults to `openruntimes-runtimes` and runtime warmup/allowlist accepts legacy `OPR_EXECUTOR_IMAGES` when `*_RUNTIMES` is not set.
+- **Drop-in server compatibility**: `/v1/health` now returns plain-text `OK` (reference behavior), while enhanced JSON stats moved to `/v1/health/stats`.
+- **API response parity**: Runtime delete now responds `200 OK` (instead of `204`) to match executor-main semantics.
+- **Response header parity**: All responses now include `Server: Executor`, matching executor-main startup request handling.
+- **Storage startup compatibility**: When `*_STORAGE_DEVICE` is unset, executor now honors `URT_CONNECTION_STORAGE` / `OPR_EXECUTOR_CONNECTION_STORAGE` DSN during storage initialization.
+
 ## [0.1.5] - 2026-02-25
 
 ### Fixed

Cargo.lock

@@ -6,7 +6,7 @@ members = [
 ]
 
 [workspace.package]
-version = "0.1.5"
+version = "0.2.0"
 edition = "2021"
 authors = ["Unified Runtimes Contributors"]
 license = "MIT"