October Product Update

Plural Telemetry, persistent filesystem, KubeCon, and more!

📊Plural Telemetry: An S3-native observability stack for metrics, logs, and traces

We're releasing Plural Telemetry, a Rust reimplementation of the open source observability stack, built directly on object storage.

Drop-in compatible. Metrics is Prometheus-compatible with remote write and OTLP ingest, Logs is Loki-compatible, and Traces is Tempo-compatible — the same APIs your collectors and dashboards already speak.
Faster and lighter. In our differential benchmarks, metrics queries run 2.5–5x faster than Mimir on about 1/18 of its CPU, traces use about a fifth of Tempo's CPU, and logs use under 40% of Loki's peak memory.
S3 all the way down. Every store keeps its data in object storage via SlateDB, with NVMe caching for hot reads — the same foundation behind Workbench Files.
Built for Kubernetes. An operator runs each store standalone or sharded, scales writers without moving data, and handles multi-tenancy plus basic or JWT auth.
Open source. Source is AGPL v3 with a commercial option; prebuilt binaries and images are MIT, and the operator and Helm charts are Apache 2.0.

Persistent workbench storage. Every workbench now gets its own filesystem, backed by the object store your Console already uses — S3, Azure, and now Google Cloud Storage.
A knowledge base for agents. Agents read and write files at /workbench in their sandbox, with built-in search and CSV, YAML, zip, and tar handling, so runbooks, exports, and reference data stay available across jobs.
Prompt attachments. Attach files to workbench, queued, and agent run prompts. They land in the workbench filesystem and are passed along to coding agents.
Files tab. Browse a workbench's files in a tree, preview them, and download or delete them right from the Console.

ServiceNow tool. Workbenches can now connect to ServiceNow, letting agents search, read, create, update, comment on, and transition incidents, change requests, problems, and catalog tasks. Agents can't delete tickets.
ServiceNow webhooks. ServiceNow can send ticket changes to Plural through a Business Rule. They arrive as issues, and any workbench with a matching webhook trigger picks them up.
Public dashboard links. Workbench dashboards can be shared through a public link, so stakeholders can see the metrics without logging into Plural.
Board and details views. Workbench jobs and alerts get board views like issues, and all three get a details view. Selecting a job shows its status, the alert or issue that triggered it, and its PRs.
Reprompt finished agent runs. Send a follow-up prompt to a completed, failed, or cancelled agent run, and Plural restores the agent's prior session — so Claude Code, Codex, Gemini, OpenCode, and Pi continue with full context instead of starting cold.

The deployment operator now decides insight priority and cluster health with clear, deterministic rules — and lets you override them.

One insight per failure. Failures roll up to their root object, so a Deployment with ten crashlooping pods shows up once, not ten times.
Reworked health score. Every failure is classed as a core cluster component or an app. Core failures like DNS, CNI, or ingress controllers can reach CRITICAL, while a broken third-party workload never turns the cluster red on its own.
Custom priority scripts. A new CustomInsightPriority resource lets you set priority for groups of objects with a sandboxed Python script, targeted by kind, namespace, or label. An annotation pins or ignores a single object.
Service context. Failing workloads deployed by Plural now link back to the service that manages them.

Stack policy stages. Binding policies for stacks can now target a specific stage — approval, evaluated after the plan, or run, evaluated before a run is created.
Finer-grained sync options. Add Prune=False to a resource's sync-options annotation to keep it when it drops out of a service's manifests. Delete=False keeps it when the service is deleted, and detach does both.

November is a big month for us on the road! Come find the Plural team at:

Stop by and say hi!