feat: if upgrade 17 -> 17 modify upgrade process

ansible/files/admin_api_scripts/pg_upgrade_scripts/initiate.sh

@@ -41,10 +41,14 @@ LOG_FILE="/var/log/pg-upgrade-initiate.log"
 
 POST_UPGRADE_EXTENSION_SCRIPT="/tmp/pg_upgrade/pg_upgrade_extensions.sql"
 POST_UPGRADE_POSTGRES_PERMS_SCRIPT="/tmp/pg_upgrade/pg_upgrade_postgres_perms.sql"
-OLD_PGVERSION=$(run_sql -A -t -c "SHOW server_version;")
+OLD_PGVERSION=$(pg_config --version | sed 's/PostgreSQL \([0-9]*\.[0-9]*\).*/\1/')
+
+# Skip locale settings if both versions are PostgreSQL 17+
+if ! [[ "$OLD_PGVERSION" =~ ^17.* && "$PGVERSION" =~ ^17.* ]]; then
+    SERVER_LC_COLLATE=$(run_sql -A -t -c "SHOW lc_collate;")
+    SERVER_LC_CTYPE=$(run_sql -A -t -c "SHOW lc_ctype;")
+fi
 
-SERVER_LC_COLLATE=$(run_sql -A -t -c "SHOW lc_collate;")
-SERVER_LC_CTYPE=$(run_sql -A -t -c "SHOW lc_ctype;")
 SERVER_ENCODING=$(run_sql -A -t -c "SHOW server_encoding;")
 
 POSTGRES_CONFIG_PATH="/etc/postgresql/postgresql.conf"
@@ -251,7 +255,12 @@ function initiate_upgrade {
     if [ -n "$IS_LOCAL_UPGRADE" ]; then
         mkdir -p "$PG_UPGRADE_BIN_DIR"
         mkdir -p /tmp/persistent/
-        echo "a7189a68ed4ea78c1e73991b5f271043636cf074" > "$PG_UPGRADE_BIN_DIR/nix_flake_version"
+	if [ -n "$NIX_FLAKE_VERSION" ]; then
+            echo "$NIX_FLAKE_VERSION" > "$PG_UPGRADE_BIN_DIR/nix_flake_version"
+        else
+            echo "a7189a68ed4ea78c1e73991b5f271043636cf074" > "$PG_UPGRADE_BIN_DIR/nix_flake_version"
+        fi
+
         tar -czf "/tmp/persistent/pg_upgrade_bin.tar.gz" -C "/tmp/pg_upgrade_bin" .
         rm -rf /tmp/pg_upgrade_bin/
     fi
@@ -394,9 +403,14 @@ function initiate_upgrade {
     rm -rf "${PGDATANEW:?}/"
 
     if [ "$IS_NIX_UPGRADE" = "true" ]; then
-        LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -c ". /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh && $PGBINNEW/initdb --encoding=$SERVER_ENCODING --lc-collate=$SERVER_LC_COLLATE --lc-ctype=$SERVER_LC_CTYPE -L $PGSHARENEW -D $PGDATANEW/ --username=supabase_admin" -s "$SHELL" postgres
+        if [[ "$PGVERSION" =~ ^17.* ]]; then
+            LC_ALL=en_US.UTF-8 LC_CTYPE=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -c ". /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh && $PGBINNEW/initdb --encoding=$SERVER_ENCODING --locale-provider=icu --icu-locale=en_US.UTF-8 -L $PGSHARENEW -D $PGDATANEW/ --username=supabase_admin" -s "$SHELL" postgres
+        else
+            LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -c ". /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh && $PGBINNEW/initdb --encoding=$SERVER_ENCODING --lc-collate=$SERVER_LC_COLLATE --lc-ctype=$SERVER_LC_CTYPE -L $PGSHARENEW -D $PGDATANEW/ --username=supabase_admin" -s "$SHELL" postgres
+        fi
     else
         su -c "$PGBINNEW/initdb -L $PGSHARENEW -D $PGDATANEW/ --username=supabase_admin" -s "$SHELL" postgres
+
     fi
 
     # This line avoids the need to supply the supabase_admin password on the old
@@ -409,6 +423,25 @@ $(cat /etc/postgresql/pg_hba.conf)" > /etc/postgresql/pg_hba.conf
         run_sql -c "select pg_reload_conf();"
     fi
 
+    TMP_CONFIG="/tmp/pg_upgrade/postgresql.conf"
+    cp "$POSTGRES_CONFIG_PATH" "$TMP_CONFIG"
+
+    # Check if max_slot_wal_keep_size exists in the config
+    if grep -q "max_slot_wal_keep_size" "$TMP_CONFIG"; then
+        # Find and replace the existing setting
+        sed -i 's/^\s*max_slot_wal_keep_size\s*=.*$/max_slot_wal_keep_size = -1/' "$TMP_CONFIG"
+    else
+       # Add the setting if not found
+       echo "max_slot_wal_keep_size = -1" >> "$TMP_CONFIG"
+    fi
+
+    # Remove db_user_namespace if upgrading from PG15
+    if [[ "$OLD_PGVERSION" =~ ^15.* && "$PGVERSION" =~ ^17.* ]]; then
+        sed -i '/^db_user_namespace/d' "$TMP_CONFIG"
+    fi
+
+    chown postgres:postgres "$TMP_CONFIG"
+
     UPGRADE_COMMAND=$(cat <<EOF
     time ${PGBINNEW}/pg_upgrade \
     --old-bindir="${PGBINOLD}" \
@@ -417,17 +450,23 @@ $(cat /etc/postgresql/pg_hba.conf)" > /etc/postgresql/pg_hba.conf
     --new-datadir=${PGDATANEW} \
     --username=supabase_admin \
     --jobs="${WORKERS}" -r \
-    --old-options='-c config_file=${POSTGRES_CONFIG_PATH}' \
+    --old-options="-c config_file=$TMP_CONFIG" \
     --old-options="-c shared_preload_libraries='${SHARED_PRELOAD_LIBRARIES}'" \
     --new-options="-c data_directory=${PGDATANEW}" \
+    --new-options="-c config_file=$TMP_CONFIG" \
     --new-options="-c shared_preload_libraries='${SHARED_PRELOAD_LIBRARIES}'"
 EOF
     )
 
     if [ "$IS_NIX_BASED_SYSTEM" = "true" ]; then
         UPGRADE_COMMAND=". /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh && $UPGRADE_COMMAND"
     fi 
-    GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND --check" -s "$SHELL" postgres
+
+    if [[ "$PGVERSION" =~ ^17.* ]]; then
+        GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND --check" -s "$SHELL" postgres
+    else
+        GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND --check" -s "$SHELL" postgres
+    fi
 
     echo "10. Stopping postgres; running pg_upgrade"
     # Extra work to ensure postgres is actually stopped
@@ -439,11 +478,24 @@ EOF
 
         sleep 3
         systemctl stop postgresql
+
+        # Additional check to ensure postgres is really stopped
+        if [ -f "${PGDATAOLD}/postmaster.pid" ]; then
+            echo "PostgreSQL still running, forcing stop..."
+            pid=$(head -n 1 "${PGDATAOLD}/postmaster.pid")
+            kill -9 "$pid" || true
+            rm -f "${PGDATAOLD}/postmaster.pid"
+        fi
     else
         CI_stop_postgres
     fi
 
-    GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND" -s "$SHELL" postgres
+    # Start the old PostgreSQL instance with version-specific options
+    if [[ "$PGVERSION" =~ ^17.* ]]; then
+        GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND" -s "$SHELL" postgres
+    else
+        GRN_PLUGINS_DIR=/var/lib/postgresql/.nix-profile/lib/groonga/plugins LC_ALL=en_US.UTF-8 LC_CTYPE=$SERVER_LC_CTYPE LC_COLLATE=$SERVER_LC_COLLATE LANGUAGE=en_US.UTF-8 LANG=en_US.UTF-8 LOCALE_ARCHIVE=/usr/lib/locale/locale-archive su -pc "$UPGRADE_COMMAND" -s "$SHELL" postgres
+    fi
 
     # copying custom configurations
     echo "11. Copying custom configurations"

ansible/vars.yml

@@ -9,9 +9,9 @@ postgres_major:
 
 # Full version strings for each major version
 postgres_release:
-  postgresorioledb-17: "17.0.1.078-orioledb"
-  postgres17: "17.4.1.028"
-  postgres15: "15.8.1.085"
+  postgresorioledb-17: "17.0.1.079-orioledb"
+  postgres17: "17.4.1.029"
+  postgres15: "15.8.1.086"
 
 # Non Postgres Extensions
 pgbouncer_release: "1.19.0"

nix/docs/testing-pg-upgrade-scripts.md

@@ -0,0 +1,115 @@
+# Testing PostgreSQL Upgrade Scripts
+
+This document describes how to test changes to the PostgreSQL upgrade scripts on a running machine.
+
+## Prerequisites
+
+- A running PostgreSQL instance
+- Access to the Supabase Postgres repository
+- Permissions to run GitHub Actions workflows
+- ssh access to the ec2 instance
+
+## Development Workflow
+
+1. **Make Changes to Upgrade Scripts**
+   - Make your changes to the scripts in `ansible/files/admin_api_scripts/pg_upgrade_scripts/`
+   - Commit and push your changes to your feature branch
+   - For quick testing, you can also edit the script directly on the server at `/etc/adminapi/pg_upgrade_scripts/initiate.sh`
+
+2. **Publish Script Changes** (Only needed for deploying to new instances)
+   - Go to [publish-nix-pgupgrade-scripts.yml](https://github.com/supabase/postgres/actions/workflows/publish-nix-pgupgrade-scripts.yml)
+   - Click "Run workflow"
+   - Select your branch
+   - Run the workflow
+
+3. **Publish Binary Flake Version** (Only needed for deploying to new instances)
+   - Go to [publish-nix-pgupgrade-bin-flake-version.yml](https://github.com/supabase/postgres/actions/workflows/publish-nix-pgupgrade-bin-flake-version.yml)
+   - Click "Run workflow"
+   - Select your branch
+   - Run the workflow
+   - Note: Make sure the flake version includes the PostgreSQL version you're testing (e.g., 17)
+
+4. **Test on Running Machine**
+   ssh into the machine
+   ```bash
+   # Stop PostgreSQL
+   sudo systemctl stop postgresql
+
+   # Run the upgrade script in local mode with your desired flake version
+   sudo NIX_FLAKE_VERSION="your-flake-version-here" IS_LOCAL_UPGRADE=true /etc/adminapi/pg_upgrade_scripts/initiate.sh 17
+   ```
+   Note: This will use the version of the script that exists at `/etc/adminapi/pg_upgrade_scripts/initiate.sh` on the server.
+   The script should be run as the ubuntu user with sudo privileges. The script will handle switching to the postgres user when needed.
+
+   In local mode:
+   - The script at `/etc/adminapi/pg_upgrade_scripts/initiate.sh` will be used (your edited version)
+   - Only the PostgreSQL binaries will be downloaded from the specified flake version
+   - No new upgrade scripts will be downloaded
+   - You can override the flake version by setting the NIX_FLAKE_VERSION environment variable
+   - If NIX_FLAKE_VERSION is not set, it will use the default flake version
+
+5. **Monitor Progress**
+   ```bash
+   # Watch the upgrade log
+   tail -f /var/log/pg-upgrade-initiate.log
+   ```
+
+6. **Check Results**
+   In local mode, the script will:
+   - Create a new data directory at `/data_migration/pgdata`
+   - Run pg_upgrade to test the upgrade process
+   - Generate SQL files in `/data_migration/sql/` for any needed post-upgrade steps
+   - Log the results in `/var/log/pg-upgrade-initiate.log`
+
+   To verify success:
+   ```bash
+   # Check the upgrade log for completion
+   grep "Upgrade complete" /var/log/pg-upgrade-initiate.log
+
+   # Check for any generated SQL files
+   ls -l /data_migration/sql/
+
+   # Check the new data directory
+   ls -l /data_migration/pgdata/
+   ```
+
+   Note: The instance will not be upgraded to the new version in local mode. This is just a test run to verify the upgrade process works correctly.
+
+## Important Notes
+
+- The `IS_LOCAL_UPGRADE=true` flag makes the script run in the foreground and skip disk mounting steps
+- The script will use the existing data directory
+- All output is logged to `/var/log/pg-upgrade-initiate.log`
+- The script will automatically restart PostgreSQL after completion or failure
+- For testing, you can edit the script directly on the server - the GitHub Actions workflows are only needed for deploying to new instances
+- Run the script as the ubuntu user with sudo privileges - the script will handle user switching internally
+- Local mode is for testing only - it will not actually upgrade the instance
+- The Nix flake version must include the PostgreSQL version you're testing (e.g., 17)
+- In local mode, only the PostgreSQL binaries are downloaded from the flake - the upgrade scripts are used from the local filesystem
+- You can override the flake version by setting the NIX_FLAKE_VERSION environment variable when running the script
+
+## Troubleshooting
+
+If the upgrade fails:
+1. Check the logs at `/var/log/pg-upgrade-initiate.log`
+2. Look for any error messages in the PostgreSQL logs
+3. The script will attempt to clean up and restore the original state
+4. If you see an error about missing Nix flake attributes, make sure the flake version includes the PostgreSQL version you're testing
+
+Common Errors:
+- `error: flake 'github:supabase/postgres/...' does not provide attribute 'packages.aarch64-linux.psql_17/bin'`
+  - This means the Nix flake version doesn't include PostgreSQL 17 binaries
+  - You need to specify a flake version that includes your target version
+  - You can find valid flake versions by looking at the commit history of the publish-nix-pgupgrade-bin-flake-version.yml workflow
+
+## Cleanup
+
+After testing:
+1. The script will automatically clean up temporary files
+2. PostgreSQL will be restarted
+3. The original configuration will be restored
+
+## References
+
+- [publish-nix-pgupgrade-scripts.yml](https://github.com/supabase/postgres/actions/workflows/publish-nix-pgupgrade-scripts.yml)
+- [publish-nix-pgupgrade-bin-flake-version.yml](https://github.com/supabase/postgres/actions/workflows/publish-nix-pgupgrade-bin-flake-version.yml)