1.6.0-community is out! #661

jotak · 2024-05-28T10:04:50Z

jotak
May 28, 2024
Maintainer

Highlights

The first thing you may notice is that we changed the versioning process: it's now going to be more tied to our downstream releases, with a suffix -community to avoid any ambiguity. So, downstream Network Observability 1.6 will match this upstream NetObserv 1.6.z-community. As you can see, the community comes first, the dowstream not being released yet. Note that the z-stream bit of the version may not always match, because we may do more releases upstream than downstream, or the other way around. But from now on, 1.x downstream == 1.x upstream. And in case you wonder, the code is the same.

Now that the boring part is said, what cool new features do we have?

FlowMetrics API

The FlowMetrics API is GA! It was previously a dev preview, and we've added a bunch of exciting things to make it ready for production, such as more filtering options, supporting dashboard creation, and also providing feedback on metrics cardinality. As a reminder, this FlowMetrics API allows you to define your own Prometheus metrics out of the flow logs in a flexible way. And whoever says metrics says alerts. It opens up a new horizon for network observability. More info about that here.

Agent filters

The eBPF agents now have many more filtering options, which allows you to reduce NetObserv footprint when you don't need to collect all the network flows. Among them, filtering by source or dest port, IP, protocol, etc. This makes NetObserv suitable on very specific observability use cases. More about that here.

Less dependent on Loki

When NetObserv started, it was fully dependent on Loki for storing network logs. Then we made it optional, in case you would be just interested in the couple of Prometheus metrics that NetObserv generates, and/or in getting the raw flows exported via Kafka or IPFIX. In 1.6 we push that even further, first thanks to the FlowMetrics API allowing you to create any metrics, which doesn't need Loki at all; and second, with a whole new feature allowing our OpenShift Console plugin to work out of Prometheus metrics, as well as Loki logs. While Loki still is necessary to see the full details of every flow logs, now the Topology and Overview views of the Console are unlocked even when you don't use Loki.

CLI

While this isn't strictly speaking part of the NetObserv operator release, we can't avoid mentioning the new CLI! This is an alternative way of using the different components of NetObserv: ideal for on-demand troubleshooting. You can get flows in a TUI (or in files), filter for whatever you want, even do packet capture and get pcap files, compatible with Wireshark. Check out the repo.

And more ...

That's not all, there's also a new subnets labelling feature, useful to identify external traffic and/or traffic to any custom workload; new pod scheduling options to better configure how NetObserv is deployed; the eBPF agents implement the new TCx hook; improved agent performances; more de-duplication of flows to decrease storage space...

Next is the full release note.