Launch gateway hardening: probes, retries, rate limits, forwarding deadlines - #868

Open

lorenzo-norcini-scale wants to merge 28 commits into

mainfrom

lorenzonorcini/launch-gateway-incident-hardening

Commits on Aug 14, 2026

fix(gateway): async healthcheck handler + explicit readiness probe timeout

lorenzo-norcini-scale
and
claude
committed
ae426ee
fix(chart): stop retrying 503 at the gateway VirtualService; make retry policy configurable

lorenzo-norcini-scale
and
claude
committed
5b083f4
fix(chart): add istio-proxy CPU request on gateway pods

lorenzo-norcini-scale
and
claude
committed
f6f90c6
[MLI-8206] Set an explicit, configurable timeout on the sync forwarder

lorenzo-norcini-scale
and
claude
committed
e5d11a2
Validate timeout_seconds at forwarder load time

lorenzo-norcini-scale
and
claude
committed
7d7b40f
fix(gateway): run async-task polls on a dedicated thread limiter

lorenzo-norcini-scale
and
claude
committed
e4cf62e
fix(db): build SQLAlchemy engines lazily, one per session kind on first use

lorenzo-norcini-scale
and
claude
committed
87d5140
feat(chart): per-pod local_ratelimit on gateway sidecars + named-tenant clamps

lorenzo-norcini-scale
and
claude
committed
e604287
feat(gateway): per-user request rate limiting at authentication, fail-open

lorenzo-norcini-scale
and
claude
committed
2dde946
docs: add async-task polling guidance (backoff + jitter, bounded outstanding set)

lorenzo-norcini-scale
and
claude
committed
198ef97
refactor: review cleanups on incident-hardening branch

lorenzo-norcini-scale
and
claude
committed
09372cc
fix(gateway): atomic rate-limit counter TTL; sample log-only warnings

lorenzo-norcini-scale
and
claude
committed
962ce90
fix(chart+gateway): Envoy 1.23 compatibility, Redis brownout breaker, DB build lock

lorenzo-norcini-scale
and
claude
committed
62bd717

Commits on Aug 17, 2026

Commits on Aug 18, 2026

Commits on Aug 19, 2026

fix(chart): remove unauthenticated tenant clamp
lorenzo-norcini-scale
committed
534095c

Navigation Menu

Uh oh!

Launch gateway hardening: probes, retries, rate limits, forwarding deadlines - #868

Launch gateway hardening: probes, retries, rate limits, forwarding deadlines#868
lorenzo-norcini-scale wants to merge 28 commits into
mainfrom
lorenzonorcini/launch-gateway-incident-hardening

Commits on Aug 14, 2026

fix(gateway): async healthcheck handler + explicit readiness probe timeout

fix(chart): stop retrying 503 at the gateway VirtualService; make retry policy configurable

fix(chart): add istio-proxy CPU request on gateway pods

[MLI-8206] Set an explicit, configurable timeout on the sync forwarder

Validate timeout_seconds at forwarder load time

fix(gateway): run async-task polls on a dedicated thread limiter

fix(db): build SQLAlchemy engines lazily, one per session kind on first use

feat(chart): per-pod local_ratelimit on gateway sidecars + named-tenant clamps

feat(gateway): per-user request rate limiting at authentication, fail-open

docs: add async-task polling guidance (backoff + jitter, bounded outstanding set)

refactor: review cleanups on incident-hardening branch

fix(gateway): atomic rate-limit counter TTL; sample log-only warnings

fix(chart+gateway): Envoy 1.23 compatibility, Redis brownout breaker, DB build lock

Commits on Aug 17, 2026

fix(chart): match rate limiter workload port

fix(gateway): address forwarding and DB review findings

fix(db): await async engine disposal

fix(chart): inbound vhost is named by Service port, verified on Istio 1.15

fix(chart): apply rate-limit actions at HTTP_ROUTE, not VIRTUAL_HOST

test: cover limiter breaker/sampling/client-cache, task bulkhead, DB kind matrix, forwarder timeout plumbing

Commits on Aug 18, 2026

fix(db): dispose expired async engines safely when a loop is already running

fix(chart): case-insensitive auth-scheme match in tenant clamp

chore(chart): regenerate istio attribute-match conditions from the live OpenAPI schema

fix(chart): resize route buckets against hardened-build capacity

fix(chart): size the poll bucket for bytes, not just requests

feat(gateway): per-tenant rate-limit decision metric

fix(gateway): enforce Celery forwarding deadline

revert(db): remove lazy engine changes

Commits on Aug 19, 2026

fix(chart): remove unauthenticated tenant clamp