Commit Graph
24 Commits
Author SHA1 Message Date
Matthew Meszaros d5fb0d5a3b feat: hand each fleet node only its own env by verifying a join-issued per-node NODE_CREDENTIAL stored hashed on fleet_nodes, render the resolved DEPLOYMENT_MODE, keep the instance credential keys and the local KMS master key off workers by brokering local KMS for them, require an https control-plane address on hosted production, and log every node env hand-out 2026-10-26 06:51:01 +01:00
Matthew Meszaros 88e4370f1f feat: merge main into the mailbox, session and input validation branch, move the resolved MAIL_ALLOW_PRIVATE_HOSTS and MAIL_TLS_INSECURE node env lines into fleetnode.RenderEnv, renumber the mailbox TLS pin migration to 000324, and keep both sides of the configuration reference 2026-10-24 12:20:12 +02:00
Matthew Meszaros 21acf00257 feat: resume a key rotation stage that died part way from its saved intent with per-phase state tokens, deploy all to only the ECS services an install runs, fail updater --check-env while another update holds the lock, keep the running image when adopting a node, judge rollout liveness on the service clock, deny the CI roles changes to their own permissions, and validate the Terraform plan class before acting on it 2026-10-24 09:34:01 +02:00
Matthew Meszaros 9c2cffae58 feat: deploy stable releases to AWS automatically with provenance checks, a migration gate, alarm-watched waves and reverse rollback, move worker and consumer hosts in canary rollouts with host-side health rollback and opt-in env refresh, rotate instance keys inside AWS through a rotator task and the rotate-keys workflow, plan and apply the AWS Terraform from CI with approval for replacements, and add the admin Keys tab and the matching runbooks 2026-10-24 09:10:52 +02:00
Matthew Meszaros cc1dc2f64c feat: answer refused realtime handshakes with a JSON code and reason, name a reused ws ticket or signed-out session, use the not_a_member slug on org joins, warn at boot when REALTIME_ALLOW_QUERY_CREDENTIALS is on, stop the admin socket reconnecting on admin_mfa_required, not_an_admin and jwt_required with a clear message, and have the seeder, updater, webhook and notification URL checks, instance health, fleet join and log mail transport say which setting applies and what to do 2026-10-23 19:53:03 +02:00
Matthew Meszaros 507c98e008 feat: preserve retry-safe locked fleet enrollment identities and trusted canonical CLI credential roots while disclosing unavailable generic runtime log sources 2026-10-10 07:55:46 +00:00
Matthew Meszaros d8e99877b2 feat: let an update finish work in flight instead of cutting it off, by having the import runner stop claiming on shutdown, hand back rows that never started without counting the claim and finish the rows it is connecting on an uncancelled context, the backend wait up to 45 seconds for those rows and for in-flight connects, the event bus finish the message being handled before it stops reading, and node units, the repository compose file and the installer's stack allow 60 seconds before a forced stop, with the grace period documented for Railway, Docker and Kubernetes 2026-09-24 18:31:06 +02:00
Matthew Meszaros b0050507e4 fix: tell a joining node a tag that exists, falling back to the published prod tag instead of a latest this project has never published, and drop the WORKER_IMAGE docs left behind by the removed push-based provisioning (#455) 2026-09-12 01:00:11 -07:00
Matthew Meszaros 510ee692ba feat: address the review on the split-deployment branch by moving the two broker routes onto their own NODE_BROKER_TOKEN so the internet-facing tracking and forms services no longer hold a credential that can open any organization's data key, refusing to presign any key outside the prefixes a node reaches, fixing IAM policies that named an alias ARN KMS never resolves in a Resource element, bounding both brokered HTTP clients because the sync loop's context never expires, no longer reporting a 403 from the object store as a missing body, and redacting the DSN and URL credentials the dry-run listing printed in clear 2026-09-10 14:19:53 +02:00
Matthew Meszaros 47ba13083e feat: make a split deployment work end to end by fixing the three defects that made an off-host node impossible to configure (nodeEnvKeys shipped S3_BUCKET and KMS_KEY_ID, which nothing reads, so an AWS-backed node silently used the default bucket and key alias; a joined consumer never received PRIMARY_DB and died at boot; and node.env was rewritten on every join with no file an operator could add to), then removing the need for cloud credentials on a node at all with brokered KMS and blob providers that renderNodeEnv hands out automatically, plus deploy/split-cloud, scripts/aws-bootstrap.sh, two fleet instance checks and the docs 2026-09-10 13:58:59 +02:00
Matthew Meszaros a5993a1ed0 feat: put the standalone-statement note where the docs said it was, shellcheck the join-script checker as well since its own disable directives are load-bearing, and stop claiming the EnvironmentFile assertion covers two render variants when only the mount list varies with the environment 2026-09-09 07:04:58 -07:00
Matthew Meszaros 9203695a41 feat: harden the join-script guard against being fooled rather than against being reformatted, matching the ensure_blob_root call on its first field after a looser regex proved satisfiable by the name appearing inside a warn string, capturing the function body without a pipeline so a renamed function reports that instead of the assertion it happened to fail, and rendering both unit variants so the no-blob one is covered too 2026-09-09 06:59:14 -07:00
Matthew Meszaros eb4e1a8bb8 feat: make the join-script checker survive its own mutation tests, matching the ensure_blob_root call on the command field so a commented-out call no longer passes, tolerating a space before the parentheses in a function definition, and capturing the function body before asserting so a renamed function reports that rather than the assertion it happened to fail 2026-09-09 06:47:56 -07:00
Matthew Meszaros 6947aca9b3 feat: make the join-script check assert on what the script renders via a new --print-unit mode, after the previous version compared a heredoc copied into the checker and stayed green when the systemd command-substitution bug was put back, and make docker_mounts pure so the unit can be rendered without creating directories 2026-09-09 06:37:56 -07:00
Matthew Meszaros 29bf9b4320 feat: cover the join script with make join-check, wired into make lint, since nothing tested the highest-consequence non-Go file in the repo and three defects reached the branch through it; and fix the four the review found in the last round, matching the fs provider alias, creating blob parents under a 0022 umask rather than the 0700 one write_config leaves set, warning instead of silently mounting a blob root the node cannot write, and validating the path right after enrolment rather than halfway through the install 2026-09-09 06:28:30 -07:00
Matthew Meszaros 03255754ff feat: read the blob settings from the env in memory so the dry run reports what a real join would do instead of a previous join's file, create the blob root 0755 rather than inheriting the 0700 umask a co-located backend cannot read, refuse a relative BLOB_FS_ROOT that docker cannot mount, and stop discarding the errors that made a failed mount look like a successful join 2026-09-09 06:12:56 -07:00
Matthew Meszaros 75a068cd07 feat: always create and mount BLOB_FS_ROOT for a node because its storage layer exits when the directory is not writable and an unmounted path is root-owned by docker, and say plainly in the script and the docs that filesystem blobs cannot serve a fleet at all since the worker has to read the body the backend wrote 2026-09-09 06:06:04 -07:00
Matthew Meszaros 4fc6d1db77 feat: put every docker mount on one line in the generated unit, because command substitution strips the trailing newline and my previous multi-line blob mount collapsed the continuation so the node restart-looped on any filesystem-blob install, and mount BLOB_FS_ROOT only when it already exists rather than creating and recursively chowning a directory a co-located bare-metal backend owns 2026-09-09 05:59:44 -07:00
Matthew Meszaros 0a08983158 feat: keep the node's write path reachable and the image reference root-owned, by granting traverse on the state directory without making it writable, re-owning image-ref on every join so a machine joined by the earlier chown-everything script converges, mounting BLOB_FS_ROOT separately when blobs live under the state directory the mount no longer covers, and renaming the reserve dialog's remaining convert-to-dedicated button 2026-09-09 05:53:28 -07:00
Matthew Meszaros 65fe82c472 feat: stop the node container being able to choose the image root runs, by giving it a dedicated uid-1000 subdirectory to write its update target into instead of making the whole state directory writable, validating the tag it asks for before the updater acts on it, and no longer recursively chowning a directory a bare-metal install also keeps its blobs in 2026-09-09 05:41:16 -07:00
Matthew Meszaros bfdbd77b04 feat: fix the eight defects the second review pass found, including three where the previous fix did not land: the bind-mounted state dir was root-owned so the node running as uid 1000 still could not write its update target, elevating the reserved-worker eviction did nothing because the rotation loop bailed on target-equals-current before the urgency was consulted, and the warmup-pool assertion was vacuous which hid that warmupPoolFor checked the subscription repo before the billing provider and answered free on a self-host install 2026-09-09 05:32:30 -07:00
Matthew Meszaros 8b83062eff feat: fix ten defects the review found, the worst being that every seeder still wrote the worker columns migration 000141 moved to fleet_nodes so make dev could not seed, that the generated systemd unit used a command substitution systemd never expands so a joined machine restart-looped while the script reported success, that the state directory was not bind-mounted so the node wrote its update target inside the container where the host timer never saw it, and that placement stopped assigning warmup pool membership so paid mailboxes silently warmed in the free pool 2026-09-09 05:21:17 -07:00
Matthew Meszaros 88992c396c feat: hand the node its env base64 encoded at join instead of as a raw JSON string, because the POSIX-sh join script has no JSON parser and its sed extraction appended a stray ","heartbeat_seconds line to node.env, and refuse a heartbeat that re-registers an existing node id under the other role, which would have left a worker's mailboxes assigned to a machine that had stopped doing worker work 2026-09-09 05:01:13 -07:00
Matthew Meszaros 435dbb522f feat: replace the worker tier/type/risk-pool/egress categories with a scored placement model and make the fleet pull-based, so a machine joins with one command, workers and consumers share one node registry with usage and liveness, nodes self-update to the version the control plane resolves, and the Hetzner provisioning, worker profiles and SSH orchestrator are removed 2026-09-09 04:54:01 -07:00