chore: add commands to seed cluster dev instance with test data

CLAUDE.md

@@ -108,6 +108,11 @@ just helm::lint                 # helm lint + full render against tools/helm/lin
 just helm::template [args]      # render the chart to stdout
 just helm::check-bump           # fail if helm-chart/ changed without a version bump
 just helm::publish              # push the chart if Chart.yaml's version is unpublished
+
+# Deployed dev instance only — takes the kube context, refuses any other cluster
+just cluster::backup <context>  # pg_dumpall to ~/hackagon-backups
+just cluster::wipe <context>    # backup, then drop + recreate the app DB; dev restarts empty
+just cluster::reseed <context>  # backup, wipe, then seed; refuses if components/backend differs from origin/main
 ```
 
 Backend listens on **:3000**, frontend on **:8081**. Dev users (Keycloak

justfile

@@ -15,6 +15,7 @@ mod codegen "./tools/just/codegen.just"
 mod clean "./tools/just/clean.just"
 mod version "./tools/just/version.just"
 mod helm "./tools/just/helm.just"
+mod cluster "./tools/just/cluster.just"
 
 [private]
 default:

tools/just/cluster.just

@@ -0,0 +1,269 @@
+set positional-arguments
+set shell := ["bash", "-cue"]
+
+root_dir := `git rev-parse --show-toplevel`
+
+# Names the chart gives a release called `hackagon` (see helm-chart/templates).
+namespace := "hackagon"
+backend_deploy := "hackagon-backend"
+backend_config := "hackagon-backend-config"
+backend_db_secret := "hackagon-backend-db"
+postgres_sts := "hackagon-postgresql"
+app_db := "hackagon"
+app_db_owner := "hackagon"
+
+# Refuse any cluster whose ingress does not serve this domain, so a recipe
+# pointed at the prod context by mistake stops before touching anything.
+dev_domain := "hackagon-dev.dscompute.ch"
+
+backup_dir := env("HOME") + "/hackagon-backups"
+local_pg_port := "15432"
+postgres_svc := "svc/{{postgres_sts}}"
+
+# Runs inside the postgres pod: finds the superuser password the bitnami image
+# was started with (env var or mounted file) so it never leaves the pod.
+pg_as_superuser := '''
+pw="${POSTGRES_POSTGRES_PASSWORD:-${POSTGRES_PASSWORD:-}}"
+if [ -z "$pw" ]; then
+  f="${POSTGRES_POSTGRES_PASSWORD_FILE:-${POSTGRES_PASSWORD_FILE:-}}"
+  [ -n "$f" ] && pw="$(cat "$f")"
+fi
+PGPASSWORD="$pw" exec "$@"
+'''
+
+default:
+    just --list --unsorted -f "{{source_file()}}"
+
+# Show usage examples for cluster commands.
+[group('cluster')]
+help:
+    #!/usr/bin/env bash
+    bold="\033[1m"
+    dim="\033[2m"
+    cyan="\033[36m"
+    reset="\033[0m"
+
+    echo ""
+    echo -e "  ${bold}Dev Cluster Commands${reset}"
+    echo -e "  ${dim}Act on the deployed dev instance. Every recipe takes the kube context${reset}"
+    echo -e "  ${dim}explicitly and refuses a cluster not serving {{dev_domain}}.${reset}"
+    echo ""
+    echo -e "  ${cyan}just cluster::backup${reset} <context>"
+    echo -e "  ${dim}pg_dumpall the dev postgres to {{backup_dir}}${reset}"
+    echo ""
+    echo -e "  ${cyan}just cluster::wipe${reset} <context>"
+    echo -e "  ${dim}Back up, then drop and recreate the app database. Keycloak is untouched.${reset}"
+    echo -e "  ${dim}The backend recreates the schema on startup, so dev comes back empty.${reset}"
+    echo ""
+    echo -e "  ${cyan}just cluster::reseed${reset} <context>"
+    echo -e "  ${dim}Back up, wipe, then run cmd/seed (backend code must match origin/main).${reset}"
+    echo ""
+    echo -e "  ${dim}e.g. just cluster::wipe sck-sit-dev${reset}"
+    echo ""
+
+# pg_dumpall the dev postgres (app and keycloak databases) to ~/hackagon-backups.
+[group('cluster')]
+backup context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    just -f "{{source_file()}}" _guard "{{context}}"
+    mkdir -p "{{backup_dir}}"
+    out="{{backup_dir}}/{{context}}-$(date +%Y%m%d-%H%M%S).sql"
+
+    discard_partial() { rm -f "$out.partial"; }
+    trap discard_partial EXIT
+
+    echo "==> Backing up to $out"
+    kubectl --context "{{context}}" -n "{{namespace}}" exec "sts/{{postgres_sts}}" -- \
+        sh -c '{{pg_as_superuser}}' sh pg_dumpall -h 127.0.0.1 -U postgres > "$out.partial"
+    mv "$out.partial" "$out"
+    echo "✓ $(du -h "$out" | cut -f1) written"
+
+# Back up, then drop and recreate the app database; dev restarts empty.
+[group('cluster')]
+wipe context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    just -f "{{source_file()}}" _confirm "{{context}}" "WIPE the app database"
+    just -f "{{source_file()}}" backup "{{context}}"
+
+    kc=(kubectl --context "{{context}}" -n "{{namespace}}")
+    replicas="$("${kc[@]}" get "deploy/{{backend_deploy}}" -o jsonpath='{.spec.replicas}')"
+    [ "${replicas:-0}" -gt 0 ] || replicas=1
+
+    restore_backend() {
+        # A backend found at 0 (e.g. left there by an interrupted run)
+        # comes back at 1: both recipes promise a running dev,
+        # and wipe needs it to recreate the schema."
+        echo "==> Scaling backend back to $replicas (it recreates the schema on startup)"
+        "${kc[@]}" scale "deploy/{{backend_deploy}}" --replicas="$replicas"
+    }
+    # The backend must be off while its database is replaced; bring it back
+    # however this ends, including when stopping it or the wipe fails halfway.
+    trap restore_backend EXIT
+
+    just -f "{{source_file()}}" _backend-down "{{context}}"
+    just -f "{{source_file()}}" _wipe-db "{{context}}"
+
+# Back up, wipe, then seed the dev fixture from this checkout.
+[group('cluster')]
+reseed context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    cd "{{root_dir}}"
+
+    just -f "{{source_file()}}" _backend-matches-main
+    just -f "{{source_file()}}" _guard "{{context}}"
+
+    kc=(kubectl --context "{{context}}" -n "{{namespace}}")
+    tmp="$(mktemp -d)"
+    pf_pid=""
+    backend_down=""
+
+    restore_backend() {
+        # Starting fresh is also what loads the casbin rows the seed wrote.
+        echo "==> Scaling backend back to $replicas"
+        "${kc[@]}" scale "deploy/{{backend_deploy}}" --replicas="$replicas"
+    }
+    # However this ends: close the tunnel, delete the temp dir, and bring the
+    # backend back if it was taken down.
+    cleanup() {
+        [ -n "$pf_pid" ] && kill "$pf_pid" 2>/dev/null || true
+        rm -rf "$tmp"
+        [ -z "$backend_down" ] || restore_backend
+    }
+    trap cleanup EXIT
+
+    # Everything that can fail without changing the cluster happens first, so
+    # a failure here leaves dev exactly as it was.
+    echo "==> Building the seed"
+    (cd components/backend && go build -o "$tmp/seed" ./cmd/seed/)
+
+    echo "==> Fetching backend config and database password from the cluster"
+    "${kc[@]}" get cm "{{backend_config}}" -o jsonpath='{.data.config\.yaml}' > "$tmp/config.yaml"
+    [ -s "$tmp/config.yaml" ] || { echo "✗ {{backend_config}} has no config.yaml." >&2; exit 1; }
+    db_password="$("${kc[@]}" get secret "{{backend_db_secret}}" -o jsonpath='{.data.password}' | base64 -d)"
+    [ -n "$db_password" ] || { echo "✗ {{backend_db_secret}} has no password." >&2; exit 1; }
+
+    replicas="$("${kc[@]}" get "deploy/{{backend_deploy}}" -o jsonpath='{.spec.replicas}')"
+    [ "${replicas:-0}" -gt 0 ] || replicas=1
+
+    just -f "{{source_file()}}" _confirm "{{context}}" "WIPE and RESEED the app database"
+    just -f "{{source_file()}}" backup "{{context}}"
+
+    # The backend stays off until the seed is done: started on the empty
+    # database it would create the tables at the same time as the seed, and it
+    # would load its casbin policy before the seed has written any.
+    backend_down=yes
+    just -f "{{source_file()}}" _backend-down "{{context}}"
+
+    # Wipe the app database, leaving keycloak untouched. The seed creates the schema.
+    just -f "{{source_file()}}" _wipe-db "{{context}}"
+
+    echo "==> Port-forwarding postgres to localhost:{{local_pg_port}}"
+    "${kc[@]}" port-forward "{{postgres_svc}}" "{{local_pg_port}}:5432" >/dev/null &
+    pf_pid=$!
+
+    # Wait up to 10s for the tunnel to open.
+    for _ in $(seq 20); do
+        nc -z 127.0.0.1 "{{local_pg_port}}" 2>/dev/null && break
+        sleep 0.5
+    done
+    nc -z 127.0.0.1 "{{local_pg_port}}" 2>/dev/null \
+        || { echo "✗ The port-forward to postgres did not open within 10s." >&2; exit 1; }
+
+    echo "==> Seeding"
+    cd components/backend
+    HACKAGON_DATABASE_HOST=127.0.0.1 \
+    HACKAGON_DATABASE_PORT="{{local_pg_port}}" \
+    HACKAGON_DATABASE_PASSWORD="$db_password" \
+        "$tmp/seed" --config-dir "$tmp/"
+    echo "✓ Seeded"
+
+# Scale the backend to 0 and wait until its pods are gone. Scaling it back up is
+# the caller's job, in an EXIT trap set before calling, so it happens even when
+# this or a later step fails halfway.
+[private]
+_backend-down context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+
+    # Guard the context, so a recipe pointed at the prod context by mistake stops before touching anything.
+    just -f "{{source_file()}}" _guard "{{context}}"
+    kc=(kubectl --context "{{context}}" -n "{{namespace}}")
+
+    backend_pods() { "${kc[@]}" get pods -o name | grep "{{backend_deploy}}-" || true; }
+
+    echo "==> Scaling backend to 0"
+    "${kc[@]}" scale "deploy/{{backend_deploy}}" --replicas=0
+
+    # Wait up to 2 minutes for its pods to be gone.
+    for _ in $(seq 60); do
+        [ -z "$(backend_pods)" ] && break
+        sleep 2
+    done
+    [ -z "$(backend_pods)" ] \
+        || { echo "✗ Backend pods still running after 2 minutes." >&2; exit 1; }
+
+# Drop and recreate the app database and run _backend-down first,
+# so the backend does not try to connect to it while it is being replaced.
+[private]
+_wipe-db context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+
+    # Guard the context, so a recipe pointed at the prod context by mistake stops before touching anything.
+    just -f "{{source_file()}}" _guard "{{context}}"
+    kc=(kubectl --context "{{context}}" -n "{{namespace}}")
+
+    echo "==> Recreating database {{app_db}}"
+    "${kc[@]}" exec "sts/{{postgres_sts}}" -- sh -c '{{pg_as_superuser}}' sh \
+        psql -h 127.0.0.1 -U postgres -v ON_ERROR_STOP=1 \
+        -c 'DROP DATABASE IF EXISTS {{app_db}} WITH (FORCE);' \
+        -c 'CREATE DATABASE {{app_db}} OWNER {{app_db_owner}};'
+    echo "✓ Database {{app_db}} is empty"
+
+# Refuse unless this checkout's backend is the code dev serves.
+[private]
+_backend-matches-main:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    cd "{{root_dir}}"
+    git fetch --quiet origin main
+    if ! git merge-base --is-ancestor origin/main HEAD; then
+        echo "✗ This checkout is behind origin/main. Rebase or merge main first." >&2
+        exit 1
+    fi
+    if ! git diff --quiet origin/main -- components/backend; then
+        echo "✗ components/backend differs from origin/main, which dev is serving." >&2
+        exit 1
+    fi
+    if [ -n "$(git status --porcelain components/backend)" ]; then
+        echo "✗ components/backend has uncommitted changes." >&2
+        exit 1
+    fi
+    echo "✓ components/backend matches origin/main"
+
+# Refuse a context whose hackagon ingress does not serve the dev domain.
+[private]
+_guard context:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    hosts="$(kubectl --context "{{context}}" -n "{{namespace}}" get ingress \
+        -o jsonpath='{.items[*].spec.rules[*].host}')"
+    if [[ " $hosts " != *"{{dev_domain}}"* ]]; then
+        echo "✗ Context '{{context}}' does not serve {{dev_domain}} (found: ${hosts:-none})." >&2
+        echo "  Refusing — this does not look like the dev instance." >&2
+        exit 1
+    fi
+
+# Guard, then make the operator type the context name back.
+[private]
+_confirm context action:
+    #!/usr/bin/env bash
+    set -euo pipefail
+    just -f "{{source_file()}}" _guard "{{context}}"
+    echo "About to {{action}} on context '{{context}}' (namespace {{namespace}})."
+    echo "Everything in the app database is lost; a backup is taken first."
+    read -r -p "Type the context name to continue: " answer
+    [ "$answer" = "{{context}}" ] || { echo "Aborted."; exit 1; }