RT @maksim_ka2: Marshaling domain structs straight into proto binary is just so convenient. Forget all the protoc, buf, endless generated c….

Why TiDB moved from Prometheus to VictoriaMetrics.

RT @PrometheusIO: We're extending PromCon EU 2025 CFP deadline to Monday! Whether you're a weekend warrior, a last-minute genius, or just n….

See also

Well-written article from Prometheus core developer about why it is better to use the classic Prometheus metrics' collection instead of jumping to hype train with OpenTelemetry-based collection.

Stuck with slow LogsQL query and do not know why it is slow and how to optimize it? Then read the following troubleshooting guide created by @func25 -

Disk read IO bandwidth is the ultimate bottleneck for the query speed when the referred data doesn't fit available RAM. The following techniques won't increase query performance in this case:. - Increasing CPU and RAM.- Optimising algorithms, which read the data from disk. The.

I'm personally not excited about all the LLM hype, but the MCP server for VictoriaMetrics provides really good results for typical observability questions (in 80% of cases :) ). Try it on your data and share the results.

gained support for Prometheus-compatible histograms thanks to @clementnuss !.

Ask the lead developer of VictoriaTraces - @ReadTheDuck - for any questions about it.

VictoriaTraces (database for traces from VictoriaMetrics) is actively developed in a dedicated repository - . Follow this repo in order to be aware of its' development. You can also try it already.

RT @VictoriaMetrics: 🚀 Did you know VictoriaMetrics offers multiple Playgrounds? 🎉.🔍 Explore how to:.Test the powerful query engine.Visuali….

An interesting comparison of Grafana Loki and VictoriaLogs on production workload from independent user - TrueFoundry -

VictoriaLogs has been moved to its own repository - Previously VictoriaLogs was co-hosted in VictoriaMetrics repo, and this was confusing for users, since VictoriaMetrics and VictoriaLogs have different release schedules.

An interesting lightning talk on why and how Spotify migrated to VictoriaMetrics.

The best availability is achieved for a distributed system where components do not know about each other, so they could continue working autonomously when the network is unavailable. This is known as cell-based architecture.

The availability of the network is usually quite high because of high availability of network switches, network media (cables) and power supply. This makes an illusion of high availability for distributed systems, which are "designed to be highly available".

The availability of a distributed system cannot exceed the availability of the network used for communication between components of the system. The distributed system enters degraded mode when the network breaks because of unavailable communications between its' components.

RT @ohmypy: Go 1.25 is coming soon, and the official release notes are pretty dry. So I prepared an interactive version with lots of examp….

RT @func25: Chapter 5 of The Anatomy of Go was released last week. What will you learn from reading it?. - What devirtualization is, and in….