It’s a packed Monday in the Postgres world. The Postgres Summit US 2026 talk schedule is officially live, four ecosystem projects shipped updates in the past week — including a pgAdmin 4 security release that patches seven CVEs and pgBackRest 2.59 with PostgreSQL 19 support — and the AI-agent era got its own managed Postgres with Tiger Data’s Ghost going GA. Here’s what Houston DBAs, data engineers, and platform teams should know.

Postgres Summit US 2026 Schedule Is Live

The United States PostgreSQL Association (PgUS) announced that the full talk schedule for Postgres Summit US 2026 is now published. The conference runs September 30 – October 2, 2026 at Convene, 555 Broadway in New York City, with the complete session list browsable on the event site.

For Houston teams, this is the flagship U.S. Postgres event of the year — the one where core developers, extension maintainers, and cloud vendors converge. Energy companies, Texas Medical Center institutions, and logistics platforms running Postgres in production send DBAs precisely for that hallway-track networking and release-cycle intel. With Summit roughly seven weeks out, now is the time to book travel and register before early-bird pricing closes.

pgAdmin 4 v9.17 Patches Seven Security Holes

The pgAdmin development team shipped 4 v9.17 with 28 bug fixes and new features — including fixes for seven security vulnerabilities (CVE-2026-17346 through CVE-2026-17351, and CVE-2026-17566). Headline fixes: an OS command injection in the MASTER_PASSWORD_HOOK feature, a command-injection path in the Import/Export Data tool that could reach COPY ... TO PROGRAM, a tool-permission bypass affecting the Query Tool and Schema Diff, and credential-cloning issues around shared servers.

Useful additions landed too: pgAdmin now records the authenticated user’s identity in the HTTP access log, the plain “View Data” action can cap fetched rows for large tables, and the Geometry Viewer supports custom XYZ tile providers. The Helm chart also gained an opt-in Gateway API HTTPRoute template.

If your Houston organization runs pgAdmin in server mode on shared infrastructure — common for energy and healthcare shops that centralize DBA tooling — this is a patch now release:

docker pull dpage/pgadmin4:latest
docker ps --filter ancestor=dpage/pgadmin4

pgBackRest 2.59 Adds PG 19 Support and WAL Archive Cleanup

The pgBackRest community released 2.59.0 on July 30 with PostgreSQL 19 support and a long list of storage and operational improvements. Standouts include a new archive-expire-before option for cleaning up the WAL archive, S3 Outposts support, per-repository backup progress in the info command, deeper verify checks that validate backup.info, and systemd notify integration.

Two operational notes: only the restore command may run as root by default (other commands now error unless you set allow-root), and there is a new optional dependency on libsystemd. A quick sanity check after upgrading:

pgbackrest version
pgbackrest --stanza=main info

For Houston teams with multi-terabyte energy or clinical datasets, the WAL archive cleanup option alone is worth the upgrade — it directly attacks the “archive disk fills up silently” failure mode.

E-Maj 5.0 and AutoBase 2.10: Tooling for Day-to-Day Ops

Two more releases rounded out the week. E-Maj 5.0.0, the fine-grained write-logging and time-travel extension, now lets non-superuser roles install and use it, eases idempotent administration scripting, and adds PostgreSQL 19 compatibility (supported range: PG 14–19). For regulated Houston industries — trading desks, clinical research, logistics audit trails — E-Maj remains one of the cleanest ways to answer “what changed, when, and who did it?”

AutoBase 2.10 takes a different angle: cluster management from the Console UI. Operators can now run switchover, failover, replica reinstallation, restart, stop/start, and cluster deletion directly from the interface, configure system parameters, and deploy clusters using local disk — with an expanded set of Ansible playbooks for maintenance automation.

AI Agents Get Their Own Postgres: Tiger Data’s Ghost Goes GA

The ecosystem milestone this week: Tiger Data launched Ghost, a managed PostgreSQL service designed specifically for AI agents. The pitch is simple: agents experiment constantly, and letting them “go wild” against shared databases gets expensive and dangerous. Ghost answers with fast forking — an exact, independent copy of a dataset in minutes or seconds via a copy-on-write storage layer — and usage-based pricing on active compute rather than per-database licenses. It supports TimescaleDB, pgvector, and PostGIS.

That is the same pattern Houston AI teams are already reaching for: a scratch database per agent working on well logs, seismic attributes, or clinical notes, instead of one shared cluster. And the scale story keeps compounding — OpenAI’s engineering team detailed how PostgreSQL powers 800 million ChatGPT users, a reminder that the world’s most AI-heavy workloads still trust Postgres underneath.

What Houston Teams Should Do This Week

• Register for Postgres Summit US 2026 (September 30 – October 2, New York City) and skim the published schedule.
• Upgrade pgAdmin 4 to v9.17 — seven CVEs, including command-injection fixes, make this urgent for server-mode deployments.
• Review the pgBackRest 2.59 release notes for the root-user and libsystemd changes before upgrading.
• Test E-Maj 5.0’s non-superuser install path on a staging cluster if you track write-audit trails.
• Evaluate a fast-fork Postgres like Ghost for your next agent experiment — one database per agent, on purpose.