.github: workflow for data-plane-controller (WIP)

.github/workflows/data-plane-controller.yaml

@@ -0,0 +1,63 @@
+name: Deploy data-plane-controller
+
+on:
+  workflow_dispatch: {}
+  push:
+    branches: [johnny/dpc-cd]
+
+env:
+  CARGO_INCREMENTAL: 0 # Faster from-scratch builds.
+
+jobs:
+  build:
+    runs-on: ubuntu-24.04
+    permissions:
+      # Permissions required of the Github token in order for
+      # federated identity and authorization to work.
+      contents: read
+      id-token: write
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          submodules: true
+          lfs: true
+
+      - uses: supabase/setup-cli@v1
+      - run: supabase start
+
+      - name: Build `data-plane-controller`
+        run: cargo build --release -p data-plane-controller
+
+      - run: mv target/release/data-plane-controller crates/data-plane-controller/
+
+      - name: Authenticate with GCP Workload Identity Federation
+        uses: google-github-actions/auth@v2
+        with:
+          service_account: [email protected]
+          workload_identity_provider: projects/1084703453822/locations/global/workloadIdentityPools/github-actions/providers/github-actions-provider
+
+      - name: Update data-plane-controller job
+        uses: google-github-actions/deploy-cloudrun@v2
+        with:
+          job: data-plane-controller
+          project_id: estuary-control
+          region: us-central1
+          source: crates/data-plane-controller/
+          timeout: 2h # Self-cancels after 1 hour, with 1 hour grace period.
+
+          env_vars: |-
+            DPC_DATABASE_CA=/etc/db-ca.crt
+            DPC_DATABASE_URL=postgresql://[email protected]:5432/postgres
+            NO_COLOR=1
+
+          secrets: |-
+            CONTROL_PLANE_DB_CA_CERT=CONTROL_PLANE_DB_CA_CERT:latest
+            DPC_GITHUB_SSH_KEY=DPC_GITHUB_SSH_KEY:latest
+            DPC_IAM_CREDENTIALS=DPC_IAM_CREDENTIALS:latest
+            DPC_SERVICE_ACCOUNT=DPC_SERVICE_ACCOUNT:latest
+            PGPASSWORD=POSTGRES_PASSWORD:latest
+            VULTR_API_KEY=DPC_VULTR_API_KEY:latest
+
+          env_vars_update_strategy: overwrite
+          secrets_update_strategy: overwrite

crates/automations/Cargo.toml

@@ -14,12 +14,12 @@ models = { path = "../models", features = ["sqlx-support"] }
 
 anyhow = { workspace = true }
 futures = { workspace = true }
+rand = { workspace = true }
 serde = { workspace = true }
 serde_json = { workspace = true }
 sqlx = { workspace = true }
 tokio = { workspace = true }
 tracing = { workspace = true }
-tracing-subscriber = { workspace = true }
 
 [dev-dependencies]
-rand = { workspace = true }
+tracing-subscriber = { workspace = true }

crates/automations/src/server.rs

@@ -227,8 +227,12 @@ async fn ready_tasks_iter(
     // If permits remain, there were not enough tasks to dequeue.
     // Sleep for up-to `dequeue_interval`, cancelling early if a task completes.
     if permits.num_permits() != 0 {
+        // Jitter dequeue by 10% in either direction, to ensure
+        // distribution of tasks and retries across executors.
+        let jitter = 0.9 + rand::random::<f64>() * 0.2; // [0.9, 1.1)
+
         tokio::select! {
-            () = tokio::time::sleep(dequeue_interval) => (),
+            () = tokio::time::sleep(dequeue_interval.mul_f64(jitter)) => (),
             _ = semaphore.clone().acquire_owned() => (), // Cancel sleep.
         }
     }

crates/data-plane-controller/Dockerfile

@@ -0,0 +1,47 @@
+FROM ubuntu:noble
+
+# Install required packages.
+RUN apt update -y \
+     && apt install --no-install-recommends -y \
+        ca-certificates \
+        certbot \
+        curl \
+        git \
+        openssh-client \
+        python3-certbot-dns-google \
+        python3-poetry \
+        python3-venv \
+     && rm -rf /var/lib/apt/lists/*
+
+# Install the `pulumi` CLI.
+RUN curl -fsSL https://get.pulumi.com/ | bash -s
+RUN ln -s /root/.pulumi/bin/pulumi /usr/local/bin/pulumi
+
+# Install the `sops` CLI.
+RUN curl -L -o /usr/local/bin/sops https://github.com/getsops/sops/releases/download/v3.9.1/sops-v3.9.1.linux.amd64 && chmod +x /usr/local/bin/sops
+
+# Copy in our local assets.
+COPY data-plane-controller /usr/local/bin/
+COPY entrypoint.sh /usr/local/bin/
+
+# AWS profile to expect in ~/.aws/credentials
+ENV AWS_PROFILE=data-plane-ops
+# GCP Service Account JSON credentials path.
+ENV GOOGLE_APPLICATION_CREDENTIALS=/etc/data_plane_controller.json
+# Disable host-key checks when cloning our git repo.
+ENV GIT_SSH_COMMAND="ssh -o StrictHostKeyChecking=no"
+
+ENV RUST_LOG=info
+
+CMD ["/usr/local/bin/entrypoint.sh"]
+
+# Example of running this container locally:
+# docker run --rm --net=host -it \
+#   -e CONTROL_PLANE_DB_CA_CERT="$(</home/johnny/Downloads/prod-ca-2021.crt)" \
+#   -e DPC_DATABASE_URL="${DATABASE_URL}" \
+#   -e DPC_GITHUB_SSH_KEY="$(</home/johnny/data_plane_controller.key)" \
+#   -e DPC_IAM_CREDENTIALS="$(</home/johnny/.aws/credentials)" \
+#   -e DPC_SERVICE_ACCOUNT="$(</etc/data_plane_controller.json)" \
+#   -e VULTR_API_KEY="${VULTR_API_KEY}" \
+#   -e PGPASSWORD=${PGPASSWORD} \
+#   foobar:latest

crates/data-plane-controller/entrypoint.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+
+set -o errexit
+set -o pipefail
+set -o nounset
+
+# Place secrets into expected file locations.
+# The SSH key in particular requires a trailing newline.
+mkdir /root/.aws
+printf '%s\n' "${CONTROL_PLANE_DB_CA_CERT}" > /etc/db-ca.crt
+printf '%s\n' "${DPC_GITHUB_SSH_KEY}" > /root/ssh_key
+printf '%s\n' "${DPC_IAM_CREDENTIALS}" > /root/.aws/credentials
+printf '%s\n' "${DPC_SERVICE_ACCOUNT}" > ${GOOGLE_APPLICATION_CREDENTIALS}
+
+# Start background ssh-agent, evaluate output to set variables, and add SSH key.
+chmod 0400 /root/ssh_key
+eval "$(ssh-agent -s)"
+ssh-add /root/ssh_key
+
+# Log out the IP from which we're running.
+echo "Current egress IP:"
+curl -s -S http://icanhazip.com
+
+# Start data-plane-controller in the background
+data-plane-controller &
+DPC_PID=$!
+
+# Start a background timer to send SIGINT after one hour.
+(
+  sleep 3600
+  kill -INT ${DPC_PID} 2>/dev/null || true
+) &
+
+# Wait for data-plane-controller to exit and surface it's status.
+set +o errexit
+wait ${DPC_PID}
+DPC_STATUS=${?}
+
+echo "data-plane-controller exited with status ${DPC_STATUS}"
+exit ${DPC_STATUS}

crates/data-plane-controller/src/lib.rs

@@ -17,19 +17,18 @@ pub struct Args {
         env = "DPC_DATABASE_URL",
         default_value = "postgres://postgres:[email protected]:5432/postgres"
     )]
-    #[serde(skip_serializing)]
     database_url: url::Url,
     /// Path to CA certificate of the database.
     #[clap(long = "database-ca", env = "DPC_DATABASE_CA")]
     database_ca: Option<String>,
     /// Number of tasks which may be polled concurrently.
-    #[clap(long = "concurrency", env = "DPC_CONCURRENCY", default_value = "2")]
+    #[clap(long = "concurrency", env = "DPC_CONCURRENCY", default_value = "1")]
     concurrency: u32,
     /// Interval between polls for dequeue-able tasks when otherwise idle.
     #[clap(
         long = "dequeue-interval",
         env = "DPC_DEQUEUE_INTERVAL",
-        default_value = "5s"
+        default_value = "10s"
     )]
     #[serde(with = "humantime_serde")]
     #[arg(value_parser = humantime::parse_duration)]
@@ -96,7 +95,7 @@ pub async fn run(args: Args) -> anyhow::Result<()> {
     }
 
     let pg_pool = sqlx::postgres::PgPoolOptions::new()
-        .acquire_timeout(std::time::Duration::from_secs(5))
+        .acquire_timeout(std::time::Duration::from_secs(30))
         .connect_with(pg_options)
         .await
         .context("connecting to database")?;

crates/data-plane-controller/src/stack.rs

@@ -29,7 +29,7 @@ pub struct DataPlane {
     pub control_plane_api: url::Url,
     pub data_buckets: Vec<url::Url>,
     pub gcp_project: String,
-    pub ssh_subnets: Vec<ipnetwork::Ipv6Network>,
+    pub ssh_subnets: Vec<ipnetwork::IpNetwork>,
     #[serde(default, skip_serializing_if = "Vec::is_empty")]
     pub private_links: Vec<AWSPrivateLink>,
     pub deployments: Vec<Deployment>,
@@ -86,7 +86,7 @@ pub struct AnsibleRole {
 
 #[derive(Debug, serde::Serialize, serde::Deserialize)]
 pub struct AnsibleHost {
-    pub ansible_host: std::net::Ipv6Addr,
+    pub ansible_host: std::net::IpAddr,
     pub ansible_user: String,
     pub host_fqdn: String,
     pub local_cert_pem: String,