▣ wi

dbcaba backup gaps: dev backup targets wrong db + gdrive backup-mirror broke 2026-05-16 (dbcaba has no auto-backup today)

Ref

#579 (#579)

Project

agent-ops

Status

backlog

Priority

high

Type

bug

Assigned

maintainer-movilba-cc-l

Created by

—

Created

2026-05-29T08:16:33.372Z

Updated

2026-05-29T08:53:35.205Z

Sub-items

Questions

Event log

• 2026-05-29T08:16:33.380Z · created
  wi cli
• 2026-05-29T08:16:45.918Z · progress
  2026-05-29: Elazar flagged 2 backup gaps via nightwatch -> pm-llmmsgsrv-cc coordinating. Confirmed from whey: backup-movilba-db.{service,timer} + backup-db-movilba-to-local-folder.sh are NOT on whey/not active here -> movilba backup runs on the GCABA /srv box (maintainer-movilba-cc-l, lezama). Dispatched read-only diagnosis to maintainer: (1) dev backup wrong-db target vs dbcaba, (2) gdrive backup-mirror breakage ~2026-05-16 root cause, (3) current movilba backup timers/last-run + is dbcaba backed up at all today, (4) exact fix + whether sh.git-edit or /srv-write (needs Elazar identity per srv auth boundary). Maintainer ETA ~5min. NO writes until Elazar GO.
• 2026-05-29T08:19:56.799Z · decision
  DIAGNOSIS (maintainer-movilba-cc-l, 2026-05-29): Gap1 REAL - /usr/local/bin/backup-movilba-dev-postgres.sh line 8 DB=movilba_dev dumps stale LOCAL pg; dev was repointed to remote dbcaba (MOBAAPI-608, swap done 05:09 today) so backup saves dead data. Gap2 (gdrive broke May16) NOT real - mirror ran clean May14-18 + today; it's just mirroring wrong content (=Gap1). Prod 'movilba' backup healthy/correct. dbcaba backup coverage today: ZERO. FIX: repoint dev backup to REMOTE pg_dump of dbcaba; script is root-owned /usr/local/bin (NOT sh.git) on lezama -> needs Elazar GO + apply by Elazar or lezama-os-cc-l + dbcaba creds. PENDING ELAZAR: (1) GO y/n, (2) creds via /root/.pgpass [recommended] vs env-in-unit. Maintainer pre-staging diff.
• 2026-05-29T08:21:19.216Z · progress
  Fix pre-staged (maintainer, apply-ready). dbcaba params confirmed from /srv/dev-movilba-web/.env.production: host=10.9.10.54 port=5432 dbname=movilba_dev (name unchanged; content prod-schema post-Gate-2) user=movilba_user; password in .env.production (storage = Elazar's call). DIFF: add DBHOST/DBPORT/DBUSER, replace 'su postgres pg_dump -d DB' with 'pg_dump -h -p -U -d DB -F c -f OUTFILE'. OUTFILE/MIRROR_DIR/rotation/logger unchanged (still *-movilba_dev.dmp). Apply options: .pgpass entry 10.9.10.54:5432:movilba_dev:movilba_user:<pw> in /root/.pgpass (600) [recommended], OR Environment=PGPASSWORD in backup-movilba-dev-db.service. Root write to /usr/local/bin -> Elazar or lezama-os-cc-l applies. AWAITING ELAZAR: GO y/n + cred method.
• 2026-05-29T08:26:36.464Z · progress
  DECONFLICT 2026-05-29: nw-whey-cc independently investigated the same nightwatch ask (it hit the whole ARO) and proposed installing a movilba backup on WHEY - WRONG HOST. Confirmed movilba backup timers run on LEZAMA (maintainer evidence: active backup-movilba-db.timer + backup-movilba-dev-db.timer there); whey unit files are sh.git copies, never installed. dbcaba fix stands = maintainer's lezama repoint. nw-whey told to stand down on whey-install. nw-whey's OTHER finding (whey gdrive bisync broken) is REAL + separate -> split to its own WI.
• 2026-05-29T08:53:35.205Z · done
  dbcaba dev backup repoint VERIFIED by maintainer-movilba-cc-l. .pgpass creds (10.9.10.54:5432:movilba_dev:movilba_user, 600); pg_dump reads pgpass only, password out of script+journal. Manual run: /var/backups/postgresql/2026-05-29-0552-movilba_dev.dmp 41M (vs old stale 2.7M) = confirmed dumping post-Gate-2 dbcaba prod-copy. gdrive mirror synced to gdrive:remotedocs/backup/lezama/postgres/movilba_dev. Gap closed.