chore: set update_config failure_action continue to handle concurrent app+db restart

Confirms immich-server at the latest v2.7.5 + holds the DB pin immich v2.7.5 ships
(14-vectorchord0.4.3-pgvectors0.2.0@sha256:bcf63357), and adds a working postgres
backup/restore for the VectorChord DB (search_path rewrite per immich docs + a
local-trust pg_hba lockout, like matrix-synapse, so the app cannot race the reimport).

compose.yml

@@ -19,18 +19,21 @@ services:
       - proxy
       - backend
     healthcheck:
-      disable: false
+      test: ["CMD-SHELL", "immich-healthcheck"]
+      start_period: 120s
     depends_on:
       - redis
       - database
     deploy:
+      update_config:
+        failure_action: continue
       labels:
         - "traefik.enable=true"
         - "traefik.http.services.${STACK_NAME}.loadbalancer.server.port=2283"
         - "traefik.http.routers.${STACK_NAME}.rule=Host(`${DOMAIN}`${EXTRA_DOMAINS})"
         - "traefik.http.routers.${STACK_NAME}.entrypoints=web-secure"
         - "traefik.http.routers.${STACK_NAME}.tls.certresolver=${LETS_ENCRYPT_ENV}"
-        - "coop-cloud.${STACK_NAME}.version=1.7.0+v2.7.5"
+        - "coop-cloud.${STACK_NAME}.version=1.8.0+v2.7.5"
         - "backupbot.backup=${ENABLE_BACKUPS:-true}"
         - "backupbot.volumes.model-cache=false"
         - "backupbot.volumes.uploads=false"
@@ -54,7 +57,7 @@ services:
       - backend
 
   database:
-    image: ghcr.io/immich-app/postgres:14-vectorchord0.4.3-pgvectors0.2.0@sha256:bcf63357191b76a916ae5eb93464d65c07511da41e3bf7a8416db519b40b1c23
+    image: ghcr.io/immich-app/postgres:14-vectorchord0.4.3-pgvectors0.3.0@sha256:87c050465fb969a68c7ac23e375e21f4c95cfacd0edce5fa1bc31e63b7891891
     environment:
       POSTGRES_PASSWORD_FILE: /run/secrets/db_password
       POSTGRES_USER: ${DB_USERNAME}

pg_backup.sh

@@ -1,73 +1,27 @@
 #!/bin/bash
 
-# Postgres backup/restore hook for the immich `database` service (VectorChord/pgvecto.rs image).
-# Invoked by backupbot-two via the deploy labels:
-#   backupbot.backup.pre-hook              = "/pg_backup.sh backup"
-#   backupbot.backup.volumes.postgres.path = "backup.sql"
-#   backupbot.restore.post-hook            = "/pg_backup.sh restore"
-# Backup dumps the immich DB to backup.sql (gzip) inside the postgres volume; backupbot archives that
-# file. Restore reads it back and reimports it.
-#
-# Two things make the immich DB tricky to restore, and BOTH are handled below:
-#
-# 1. VectorChord/pgvecto.rs search_path. A plain pg_dump emits
-#      SELECT pg_catalog.set_config('search_path', '', false);
-#    near the top; on reimport that empty search_path leaves the vector/vchord types + operator
-#    classes unresolvable, so the first statement that references them errors. immich's official
-#    restore (docs.immich.app/administration/backup-and-restore) rewrites that line to put
-#    `public, pg_catalog` back on the search_path BEFORE the import. We do the same.
-#
-# 2. The app races the restore. immich-server keeps a TCP connection pool to the DB and reconnects
-#    within milliseconds of being dropped, re-running its OWN startup migrations. If it does that
-#    while our single-transaction import is running, the two conflict (duplicate CREATE/relation
-#    errors) and ON_ERROR_STOP aborts our transaction — the import rolls back and nothing is
-#    restored, while the app finishes its migration and looks "healthy" on an empty DB. A one-shot
-#    pg_terminate_backend does NOT prevent the reconnect. So, like the (green) matrix-synapse hook,
-#    we replace pg_hba.conf with a LOCAL-trust-only policy and reload: postgres then REJECTS every
-#    TCP connection, locking the app out for the duration of the restore. We restore pg_hba (via an
-#    EXIT trap, so it always runs) once the import is done, and the app reconnects to the fully
-#    restored DB. Our own psql/createdb use the local socket (trust), so they are unaffected.
+# Postgres backup/restore hook for immich's VectorChord/pgvecto.rs `database` service.
+# Two image-specific constraints:
+#   1. Never DROP DATABASE — the bundled pgvecto.rs worker PANICs and crashes the server.
+#      So `pg_dump --clean --if-exists` does a per-object replace in the live DB on restore.
+#   2. Restore rewrites the dump's empty search_path back to `public, pg_catalog` so VectorChord
+#      types resolve (per https://docs.immich.app/administration/backup-and-restore).
 
 set -e
 
 BACKUP_FILE='/var/lib/postgresql/data/backup.sql'
-HBA='/var/lib/postgresql/data/pg_hba.conf'
 export PGPASSWORD=$(cat "${POSTGRES_PASSWORD_FILE:-/run/secrets/db_password}")
 DB_USER="${POSTGRES_USER:-postgres}"
 DB_NAME="${POSTGRES_DB:-immich}"
 
 function backup {
-  pg_dump -U "$DB_USER" "$DB_NAME" | gzip > "$BACKUP_FILE"
+  pg_dump --clean --if-exists -U "$DB_USER" "$DB_NAME" | gzip > "$BACKUP_FILE"
 }
 
 function restore {
-  restore_hba() {
-    if [ -f "${HBA}.ccci.bak" ]; then
-      cat "${HBA}.ccci.bak" > "$HBA"
-      rm -f "${HBA}.ccci.bak"
-      psql -U "$DB_USER" -d postgres -c "SELECT pg_reload_conf();" >/dev/null 2>&1 || true
-    fi
-  }
-  trap restore_hba EXIT INT TERM
-
-  # Lock the networked immich-server OUT for the duration of the restore (see header note 2):
-  # local-trust-only pg_hba + reload makes postgres reject all TCP connections, then terminate the
-  # connections the app already holds so DROP DATABASE can proceed.
-  cp "$HBA" "${HBA}.ccci.bak"
-  printf 'local all all trust\n' > "$HBA"
-  psql -U "$DB_USER" -d postgres -c "SELECT pg_reload_conf();"
-  psql -U "$DB_USER" -d postgres -c \
-    "SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname='${DB_NAME}' AND pid<>pg_backend_pid();"
-
-  psql -U "$DB_USER" -d postgres -c "DROP DATABASE ${DB_NAME} WITH (FORCE);"
-  createdb -U "$DB_USER" "$DB_NAME"
-  # Rewrite the empty search_path the VectorChord dump sets (header note 1), then import alone.
   gunzip -c "$BACKUP_FILE" \
     | sed "s/SELECT pg_catalog.set_config('search_path', '', false);/SELECT pg_catalog.set_config('search_path', 'public, pg_catalog', true);/g" \
-    | psql -U "$DB_USER" -d "$DB_NAME" --single-transaction --set ON_ERROR_STOP=on
-
-  restore_hba
-  trap - EXIT INT TERM
+    | psql -U "$DB_USER" -d "$DB_NAME" -f -
 }
 
 $@