Key details

  1. pg_vault_tde 1.7.2 was announced October 5, 2026 by Miriade.
  2. The release adds v5 on-disk tuple layout to address UPDATE crashes.
  3. Fixes cover TOAST crashes, unreadable all-NULL rows, wrong tde_btree range/ORDER BY results and key-rotation data corruption.
  4. Encrypted tables must be rewritten with VACUUM FULL after upgrading; old v4 rows can crash on UPDATE.
  5. WAL resource-manager ID changes 128→161; clean shutdown required.
  6. Clusters using toast_custom_rmgr must upgrade primary and standbys together; no rolling upgrade.
  7. The SQL extension version remains 1.7; pg_vault_tde_build_version() distinguishes binary builds.

What builders should take away

  1. Inventory pg_vault_tde deployment and current binary build before changing packages.
  2. Back up and test the upgrade on a copy of real encrypted data, including key rotation and query plans.
  3. Plan VACUUM FULL for every encrypted table, including lock/disk requirements and downtime.
  4. If toast_custom_rmgr is active, plan a coordinated primary/standby shutdown and upgrade rather than a rolling deployment.
  5. Verify the resulting build version and run UPDATE, TOAST, ORDER BY and key-rotation regression tests.

What changed

Maintainer Miriade announced pg_vault_tde 1.7.2 on October 5, 2026. It addresses crashes involving indexed variable-length columns and encrypted TOAST values, a case in which an all-NULL row could make a table unreadable, and incorrect answers from the tde_btree index on range, ORDER BY and merge-join operations. It also addresses data loss and corruption bugs around online key rotation, wallet KEK rotation and vault-to-wallet migrations. A new v5 on-disk tuple format and a WAL resource-manager ID change impose mandatory upgrade steps. The extension SQL version stays at 1.7, while pg_vault_tde_build_version() distinguishes the binary patch.

Why it matters

For operators encrypting database content with this extension, the release is simultaneously an urgent correctness fix and a planned-maintenance event. The vendor warns that old v4 rows remain readable but can crash on UPDATE after installing the patch, so merely replacing the binary is insufficient. VACUUM FULL rewrites every encrypted table into the v5 layout, with potential lock, disk-space and downtime implications. The WAL ID change also rules out a rolling upgrade for clusters with toast_custom_rmgr enabled. This is not a PostgreSQL core flaw or a general PostgreSQL encryption feature; it affects users of this specific extension.

Several bugs threaten data correctness, not just uptime

The release fixes segfaults during UPDATE of indexed variable-length encrypted columns and on TOAST threshold crossings. It also prevents tde_btree from answering unsupported range and ordering operations that previously could return incorrect rows, and addresses corruption/data-loss paths in key rotation and migrations. The vendor does not describe active exploitation; these are documented defects in the extension.

VACUUM FULL is part of the upgrade contract

After installing 1.7.2, the vendor instructs operators to run VACUUM FULL on every encrypted table. Without rewriting old v4-format rows to v5, UPDATE can still crash. VACUUM FULL is a heavy rewrite operation; assess locks, extra disk headroom, replication and backup/restore windows before scheduling it.

Some clusters cannot roll through this patch

The WAL resource-manager ID changes from 128 to registered ID 161. A clean shutdown is required, and when toast_custom_rmgr is enabled primary and standby servers must be upgraded together. Rolling upgrades in that configuration are explicitly unsupported.

Verify the binary patch version

The extension version number reported at the SQL layer remains 1.7 for this binary patch. The new pg_vault_tde_build_version() function distinguishes builds, so inventory tools should not rely solely on the extension's SQL version.

What to watch next

  • Whether follow-up patches refine the on-disk migration process.
  • Whether maintainers publish more detailed exposure guidance for data-corruption cases.
  • Compatibility reports for specific PostgreSQL versions and replication setups.

Still unclear

  • The public release note does not provide affected-version matrix beyond the 1.7.2 fix.
  • The note describes defects and data-loss risk, not evidence of malicious exploitation.

Sources

Direct reading behind this dossier.

2 sources

Discussion

Discussion is reader-contributed. Comments are not part of the BTN dossier or its editorial evidence.

0 visible comments

Join the discussion

Keep comments useful and relevant. Reader contributions may be moderated and are not BTN editorial evidence.

Sign in to comment