The auto-detect read PUID/PGID from /data/db ownership (UID 102 in pre-PUID images), causing all upgrading users to run as UID 102 instead of the expected UID 1000. This broke host-side access (SSH, WinSCP), made existing DATA_DIR files unwritable, and failed comskip on recordings created before the change.
DATA_DIRS chown failures (NFS, SMB/CIFS, FUSE) crashed the container under set -e. The old image never chowned these directories, so users with external mounts had a working setup that broke on upgrade with zero config changes.
- Collect mkdir and chown failures per-directory instead of crashing
- Emit a single consolidated warning listing all affected directories with current ownership and remediation steps
- Container continues to start; Django reports at runtime if it cannot write to a specific directory
- Local volumes are unaffected: chown still runs and succeeds silently
- replaced $_pg_tmp_port with $POSTRGES_PORT in migration logic. Works because "pg_ctl stop -w" is run before entrypoint.sh restarts postgres. Verified that this does not cause a race condition.
- Correcting unquoted environment variables to prevent inappropriate variables from causing execution inconsistencies (still unquoted in comments and prints)
PostgreSQL now runs as the PUID/PGID user ($POSTGRES_USER) instead of the internal postgres system user (UID 102). This fixes container startup failures when PUID/PGID is set, caused by chown permission errors on restricted filesystems (NFS root_squash, CIFS) and UID collisions with the postgres system user.
Changes:
- Run all PostgreSQL operations (initdb, pg_ctl, psql) as $POSTGRES_USER
- Auto-detect PUID/PGID from existing data owner when not explicitly set
- Validate PUID/PGID (reject zero, non-numeric values) before startup
- Migrate existing data ownership with sentinel-based skip optimization
- Use trust auth for local Unix sockets, md5 for network connections
- Add promote_app_role() and ensure_app_database() as idempotent startup guarantees that handle fresh installs, upgrades, and PUID changes
- Preserve postgres role as superuser for rollback compatibility
- Centralize /data/db ownership in 02-postgres.sh (sentinel-aware)
- Add integration test suite (20 scenarios) covering fresh installs, upgrades, restarts, PUID changes, UID collisions, bind mounts, modular mode, PG major upgrades, and end-to-end web UI verification
- Fix Postgres version check failing with restricted DB users (use $POSTGRES_DB instead of hardcoded 'postgres')
- Fix DVR recording broken in modular mode (respect DISPATCHARR_PORT instead of hardcoding 9191)
- Remove flushdb() from wait_for_redis.py to prevent Redis data loss on container restart
- Add DISPATCHARR_PORT to celery environment in docker-compose.yml
- Add depends_on health conditions for proper service startup ordering
- Add extra_hosts for host.docker.internal resolution on Linux
- Harden celery entrypoint with timeouts for JWT wait (120s) and migration wait (300s)
- Replace fragile showmigrations grep with migrate --check
- Add unit tests for DVR port resolution and flushdb removal regression
Implements automatic version validation for external databases in modular mode to ensure compatibility.
Changes:
- Added check_external_postgres_version() in 02-postgres.sh
- Integrated version check in entrypoint.sh after database connection
- Enforces minimum PostgreSQL version matching DispatcharrBase
- Allows newer versions with forward compatibility notice
- Rejects older versions with clear upgrade instructions
The version requirement automatically scales when DispatcharrBase is updated, requiring no manual maintenance. Only applies to modular deployments using external databases; AIO deployments are unaffected.
Tested and verified correct behavior with PostgreSQL v17, v18, and v16
Extend ensure_utf8_encoding() to support both internal and external PostgreSQL.
- Modular mode: Uses TCP connection with PGPASSWORD authentication
- AIO mode: Uses Unix socket as postgres user
- Explicitly set database owner during recreation (fixes missing --owner flag)
- Conversion logic (dump/drop/recreate/restore) works for both AIO and modular modes
- Skip internal PostgreSQL setup entirely when DISPATCHARR_ENV=modular
The script was creating a group named 'dispatch' (hardcoded) while
nginx.conf expected a group matching POSTGRES_USER (e.g., 'dispatcharr').
Changes:
- Use $POSTGRES_USER for group name instead of hardcoded 'dispatch'
- Make sed command more robust to match any existing user directive
This ensures the group name matches the user name, fixing the error:
"getgrnam("dispatcharr") failed in /etc/nginx/nginx.conf:1"
Fixes#877
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
