In 2025, the Percona Operator for PostgreSQL put most of its vitality into the issues that matter when PostgreSQL is working inside actual Kubernetes clusters: predictable upgrades, safer backup and restore, clearer observability, and fewer surprises from picture and HA model drift.
Backups and restores received extra resilient and extra controllable
In March, Operator 2.6.0 concentrated closely on backup and restore habits, an space the place many manufacturing points are likely to floor.
Retry logic was added so transient community points or quick timeouts wouldn’t instantly trigger backup jobs to fail and restart from scratch. Restore habits grew to become extra configurable, permitting groups to fine-tune how WAL information are dealt with throughout restoration with out modifying backup repositories straight.
An experimental cleanup finalizer was additionally launched for groups that need computerized backup elimination when clusters are deleted, whereas conserving it elective and clearly scoped.
These adjustments are usually not flashy, however they aim the failure modes that present up in incident critiques: backups that silently cease, restores that stall, and workflows that solely work when every part is ideal.
Observability moved nearer to “works throughout actual org shapes”
Operator 2.7.0 on July 18, 2025, added native PMM 3 assist whereas conserving PMM 2 compatibility. It additionally added methods to call clusters in monitoring, in order that multi-region, multi-namespace, and multi-tenant setups stay comprehensible in dashboards.
The identical launch leaned into sensible operator controls: deeper Patroni configuration override choices for troubleshooting, and enhancements round consumer schema habits and elective public schema entry for non-superusers, which reduces friction when purposes assume “default PostgreSQL” habits.
Alongside these enhancements, the workforce frolicked validating pg_tde as a customized extension in actual operator-managed environments and revealed steerage on tips on how to use it safely. These experiments have been an necessary studying part. They highlighted each what works and the place operational complexity creeps in. Primarily based on that have, it’s now clear that encryption at relaxation wants to maneuver past customized wiring and towards extra cloud native, operator-driven integration.
Fewer transferring elements within the foundations
By late within the yr, with Operator 2.8.0, a number of foundational selections have been tightened.
The Operator standardized on Patroni 4 for top availability, decreasing legacy complexity and ambiguity. PostgreSQL photos transitioned to official Percona builds with express model assist, enhancing predictability throughout patching and automation.
Help for big pages made it simpler to run memory-intensive workloads effectively, whereas expanded S3 compatibility for customized extensions helped groups in nonstandard or restricted storage environments.
Standing reporting additionally improved with the addition of observedGeneration, making it clearer when configuration adjustments had really been utilized.
Extra milestones: PostgreSQL 18 assist and built-in pgvector
Operator 2.6.0 launched built-in pgvector within the PostgreSQL photos utilized by the Operator. This eliminated the necessity for customized extension set up steps and made it considerably simpler for groups to allow vector search capabilities in a predictable, operator-managed means.
PostgreSQL 18 assist arrived later, beginning with Operator 2.8.1. With that launch, groups may start testing and adopting PostgreSQL 18 utilizing the Operator, with packaging and defaults aligned to newer upstream expectations and with out requiring adjustments to present deployment workflows.
What’s subsequent
Looking forward to Q1 2026, the main focus shifts from particular person options to deeper, opinionated integration that displays how manufacturing groups really run PostgreSQL on Kubernetes. One concrete instance is making pg_tde a first-class citizen within the operator expertise. It shouldn’t be one thing customers need to wire in as a customized extension, however as a substitute a built-in functionality that aligns with expectations already set by different databases. PostgreSQL could not have native TDE as we speak, however Percona doesn’t method this drawback in isolation. Drawing on years of expertise with MySQL and MongoDB, we perceive how infrastructure groups count on encryption, key administration, and Kubernetes workflows to behave, and we’re well-positioned to use these classes to PostgreSQL.
Past TDE, extra is to come back. Decreasing main model improve downtime with a transparent path towards near-zero downtime upgrades is an energetic space of labor. We’re additionally investing in OIDC primarily based authentication, sooner availability of minor releases shortly after upstream, and smoother operational workflows total. One other necessary aim is to decrease friction for customers coming from different ecosystems, together with making it simpler for groups at present utilizing Crunchy Knowledge emigrate to Percona, if and after they select to.
Should you take one factor away from this wrap up, let or not it’s this. The operator is more and more formed round manufacturing habits, not simply function checklists. That route is deliberate. It solely works as a result of customers floor actual world sharp edges, and contributors flip that suggestions into code, documentation, and disciplined releases. What’s subsequent is much less about including yet one more function and extra about making PostgreSQL on Kubernetes predictable, safe, and boring in the absolute best means.
