#!/usr/bin/env bash # Bash equivalent of migrate-db-naming.ps1 for start-server.sh (Linux/macOS # hosts, no PowerShell dependency). Keep both in sync; start-server.bat uses # the .ps1 version instead. # # One-time, opt-in migration of the Postgres role/database name inside an # existing data volume from "telesrv" to "owpengram". # # docker-compose.yml's POSTGRES_USER/POSTGRES_DB env vars only take effect # when Postgres bootstraps a brand-new, empty data volume -- they're silently # ignored against a pre-existing volume, so changing the compose file alone # does nothing for an install that already has data on disk. This script # talks to the already-running Postgres container directly (run it after # "docker compose up" plus a readiness wait, not before) and renames the # role/database in place via ALTER ROLE / ALTER DATABASE. # # The existing password is left untouched -- renaming doesn't change it. Only # the TELESRV_POSTGRES_DSN line in .env is patched (never any other line, and # never if that line was already customized away from the plain telesrv # defaults) so the app can still connect afterwards. # # Prints nothing when there's nothing to do. Prompts interactively at most # once; the decision is cached in a state file next to .secrets/.public_ip, so # this is never asked again. Declining is permanent: the install keeps the # "telesrv" role/database name forever. set -euo pipefail CONTAINER_NAME="${1:-owpengram-postgres}" ENV_FILE="${2:-.env}" STATE_FILE="${3:-.db_naming}" info() { echo "$@" >&2; } pg_role_exists() { local role="$1" local out out="$(docker exec "$CONTAINER_NAME" psql -U telesrv -d postgres -tAc "SELECT 1 FROM pg_roles WHERE rolname='${role}'" 2>/dev/null || true)" [[ "$out" == *1* ]] } state="" [[ -f "$STATE_FILE" ]] && state="$(tr -d '[:space:]' < "$STATE_FILE")" if [[ "$state" == "owpengram" || "$state" == "telesrv" ]]; then exit 0 fi old_exists=false new_exists=false pg_role_exists "telesrv" && old_exists=true pg_role_exists "owpengram" && new_exists=true if [[ "$old_exists" == false || "$new_exists" == true ]]; then # Nothing to migrate: either already "owpengram"-named (fresh install or # already renamed), or connecting as "telesrv" didn't work at all (custom # credentials already in place) -- leave it alone either way. echo -n "owpengram" > "$STATE_FILE" exit 0 fi if [[ ! -t 0 ]]; then info "[cfg] Postgres role/database still named 'telesrv' but running non-interactively; keeping old naming for now" exit 0 fi info "" info "== Postgres role/database naming ==" info "The Postgres role and database inside this install's data volume are still" info "named 'telesrv' (renaming docker-compose.yml alone doesn't affect existing" info "data). Renaming them to 'owpengram' keeps everything consistent. Recommended." read -rp "Rename Postgres role/database from 'telesrv' to 'owpengram' now? [Y/n] " reply if [[ "$reply" =~ ^[Nn] ]]; then echo -n "telesrv" > "$STATE_FILE" info "[cfg] keeping 'telesrv' Postgres role/database naming (won't ask again)" exit 0 fi info "[cfg] renaming Postgres role/database: telesrv -> owpengram" # Postgres refuses "ALTER ROLE RENAME" (session user cannot be # renamed), so telesrv can't rename itself. Do the database rename directly # as telesrv, then use a short-lived second superuser to rename the telesrv # role, then drop that temporary role from the now-renamed "owpengram" role's # own session. Each step is a separate connection/transaction on purpose: # combining the ALTER DATABASE and ALTER ROLE in one multi-statement call # rolls the database rename back too when the role rename fails (Postgres # treats a semicolon-separated -c string as one implicit transaction). if ! docker exec "$CONTAINER_NAME" psql -U telesrv -d postgres -c \ "ALTER DATABASE telesrv RENAME TO owpengram; CREATE ROLE _telesrv_migrate SUPERUSER LOGIN PASSWORD 'telesrv_migrate';" \ >/dev/null 2>&1; then info "[ERROR] failed to rename Postgres database / create temporary migration role" exit 1 fi if ! docker exec "$CONTAINER_NAME" psql -U _telesrv_migrate -d owpengram -c \ "ALTER ROLE telesrv RENAME TO owpengram;" >/dev/null 2>&1; then info "[ERROR] failed to rename Postgres role" exit 1 fi if ! docker exec "$CONTAINER_NAME" psql -U owpengram -d owpengram -c \ "DROP ROLE _telesrv_migrate;" >/dev/null 2>&1; then info "[WARN] renamed role/database but failed to drop the temporary _telesrv_migrate role - remove it by hand" fi echo -n "owpengram" > "$STATE_FILE" info "[ok] Postgres role/database renamed to 'owpengram'" if [[ -f "$ENV_FILE" ]]; then if grep -qE '^TELESRV_POSTGRES_DSN=postgres://telesrv:[^@]*@[^/]+/telesrv(\?.*)?$' "$ENV_FILE"; then sed -i -E 's#^TELESRV_POSTGRES_DSN=postgres://telesrv:([^@]*)@([^/]+)/telesrv(\?.*)?$#TELESRV_POSTGRES_DSN=postgres://owpengram:\1@\2/owpengram\3#' "$ENV_FILE" info "[ok] updated the TELESRV_POSTGRES_DSN line in ${ENV_FILE} (only that one line)" else info "[WARN] TELESRV_POSTGRES_DSN in ${ENV_FILE} doesn't match the plain telesrv defaults - update it by hand:" info " TELESRV_POSTGRES_DSN=postgres://owpengram:@:/owpengram?sslmode=disable" fi else info "[WARN] ${ENV_FILE} not found - if you create one, point TELESRV_POSTGRES_DSN at the 'owpengram' role/database" fi