Askr documentation
Askr is a standalone, memory-safe PHP application server written in Rust. It embeds the PHP interpreter in-process (no FastCGI, no FPM), serves it from a memory-safe hot path, and — in worker mode — boots your app once and serves many requests against it, eliminating per-request framework bootstrap.
Version 1.4.14. Production target is Linux; development also works on macOS.
Start here
| Guide | What it covers |
|---|---|
| Architecture | How Askr works: embedding, non-ZTS process-per-core, the worker loop, request lifecycle, TLS, recycling, reload, the admin plane. |
| Installation | Start here — step by step from nothing to a served Laravel app (Docker, tarball or source) |
| Building | Building libphp (macOS & Ubuntu) and the askr binary; the extension matrix. |
| Releasing | Maintainer checklist: version bump, tag, and verifying the release actually reached users |
| Configuration | The askr.toml reference, CLI flags, and environment variables. |
| CLI reference | Every command and flag (serve, doctor, config-check). |
| Hosting multiple domains | Virtual hosts ([[site]]), redirects (www→apex, http→https), multi-domain TLS. |
| Stability & compatibility | The 1.0 compatibility contract: stable surfaces + deprecation policy. |
| Laravel setup | Recommended Laravel guide — composer require kwhorne/askr-laravel, .env, runner scripts, queue/scheduler/broadcasting, durable L2, production checklist. |
| Worker mode | Boot-once-serve-many, the Laravel worker script, per-request state reset, writing your own worker. |
| Power features | Response cache + tag invalidation, coalescing, Pusher WS, askr_defer, CoW autoscaling, record/replay, fork test runner. |
| Auto-TLS (ACME) | Obtain + renew Let's Encrypt certs over HTTP-01 (--acme) — no proxy. |
| Hardening / sandbox | seccomp no-exec + Landlock filesystem sandbox (--sandbox, Linux). |
| Docker | Official multi-arch GHCR image — one container replaces app+nginx+redis+queue+cron. |
| Benchmarks | Reproducible comparison vs FrankenPHP, PHP-FPM+nginx and RoadRunner — and the PHP-vs-I/O split that shaped the roadmap. |
| io_uring core (plan) | Design notes for a Linux io_uring I/O core. Deprioritised — benchmarks show PHP is ~99.5% of request time, so I/O isn't the bottleneck. |
| CoW template | Boot once, fork workers (copy-on-write) — ~ms warm respawn + shared memory (experimental). |
| Shared cache | In-binary cache, atomic counters and rate limiting (no Redis); the Laravel driver. |
| Broadcasting | Live updates to browsers via SSE + askr_broadcast() (no Reverb/Pusher). |
| Storage backends | L1 shared memory + L2 SQL Anywhere: durable, replicated, multi-box cache/queue/pub-sub (epic elyra-2). |
| Admin dashboard | The built-in status/reload/metrics API and web dashboard. |
| Observability | Ship per-request logs to ElyraSQL / any MySQL-wire database (--features observ, ASKR_OBSERV_DSN) and query them with SQL. |
| Deployment | Production: systemd, TLS, zero-downtime reload, recycling, scaling, hardening. |
| Upgrading | How to upgrade and roll back, what's worth adopting at each version, and an honest list of what can bite you. |
| Ubuntu setup | Recommended production install on Ubuntu (release tarball, systemd, TLS, tuning). |
| Maintenance | After it's running: the 30-second check, deploys, clearing the three caches, certificates, log rotation, backups, and the mistakes already made in production. |
60-second tour
Install a self-contained release (Linux x86_64 / arm64) and serve a Laravel app:
VER=v1.4.14; ARCH=$(uname -m)
curl -fsSLO https://github.com/kwhorne/askr/releases/download/$VER/askr-${VER#v}-linux-$ARCH.tar.gz
tar xzf askr-${VER#v}-linux-$ARCH.tar.gz && cd askr-${VER#v}-linux-$ARCH
./askr-run.sh doctor
ASKR_APP_BASE=/var/www/app ./askr-run.sh serve \
--root /var/www/app/public \
--worker-script examples/laravel-worker.php \
--workers "$(nproc)" --tls-self-signed --admin 127.0.0.1:9000
Production setup (systemd, TLS, hardening): Ubuntu setup. Then Maintenance for the operating side. Building from source: Building.
What works today (1.4.14)
- Embedded PHP (non-ZTS) running real Laravel 12, ~9× the per-request/FPM model
- Multi-core via one worker process per core on a shared listen socket
- Worker mode (Octane-style) with per-request state reset — no bleed
--paranoidstate-bleed detector — is your app worker-safe?- CoW template (
--cow) — boot once, fork workers for ~ms warm respawn (experimental) - Queue workers + scheduler + sidecars in the same binary (no Horizon/cron)
- Shared cache / sessions / locks / job queue (
askr_cache_*,askr_queue_*+ Laravel drivers) — fully replaces Redis on a single box - Broadcasting — SSE +
askr_broadcast(), plus a Pusher-compatible WebSocket (--pusher, drop-in Reverb with auth) - Response cache with tag invalidation, ESI fragment assembly,
PURGE/BAN,stale-if-error, per-path[[cache.rule]], request coalescing,askr_defer()post-response work — and it survives restarts - Fleet-wide rate limiting before PHP wakes up, and canary deploys that abort and drain themselves on regression
- Multipart uploads (
$_FILES) + response compression (br/gzip) - Record & replay failing requests (
--record-errors/askr replay) - Graceful recycling + auto-respawn + crash resilience
- TLS (rustls) + HTTP/2;
--tls-self-signedfor dev; auto-TLS via ACME (--acme) - Hardening (
--sandbox, Linux): seccomp no-exec + Landlock write-restriction - Zero-downtime rolling reload on
SIGHUP, with optional canary - Request hardening: body-size limit (413), HEAD, GET/POST
- Typed
askr.tomlconfig +config-check - Built-in admin dashboard + API (status, reload, live metrics)
- In-process metrics — PHP-vs-I/O split, latency histogram, per-worker RSS
askr doctorpre-flight checks
Not yet
HTTP/3 (QUIC) and OpenTelemetry trace export. The per-core io_uring core is deprioritised: our benchmarks show PHP execution is ~99.5% of request time, so an I/O-syscall optimisation would move ~0.5% — the engine, not I/O, is the ceiling.