Announcing Istio 1.7
Istio 1.7 release announcement.
We are pleased to announce the release of Istio 1.7!
CHANGE NOTES
Get a detailed list of what's changed.
BEFORE YOU UPGRADE
Things to know and prepare before upgrading.
DOWNLOAD
Download and install this release.
DOCS
Visit the documentation for this release.
Istio’s great community
As with all of our releases, Istio 1.7 was a community effort. 200 people across over 40 companies contribute to Istio. We’d like to thank our fantastic community for their ongoing efforts: it is because of our amazing community that Istio is able to make so many improvements, quarter after quarter.
About Istio 1.7
This release continues to navigate in the direction outlined in our roadmap post, improving usability, security, reliability, and especially improving on the VM (non-Kubernetes) use case.
Here are some highlights for this release:
Security enhancements
We made sure that destination rule certificates get the full benefits of secure secret distribution with SDS (especially automatic rotation), even if they are mounted as files. This is an important security best practice.
The above item applies to Gateway pods. It is now possible for Egress Gateways that do TLS/mTLS origination to provision client certificates as secrets.
We improved Trust Domain Validation to validate TCP traffic as well.
Previously only HTTP traffic was validated. Trust Domain Validation now also supports trustDomainAliases
in the MeshConfig
resource.
ECC cryptography is helpful for providing high security while being highly efficient. We added the ability to communicate to your certificate authority using ECC.
An important part of best-practice security is to not run a process with more permissions than it needs, e.g. to protect against confused deputy attacks. As such, we modified Gateway deployments to run as non-root, by default.
If you were using source principal-based security policies, there was a bug with the Istio Gateway and mTLS that could have caused them to not be respected. This got fixed.
Ease of use improvements
A big part of making systems like Istio easy to use is in their “day 2” usage, especially in their ability to help you see potential problems. We’re adding to the ability of the very useful istioctl analyze tool:
For frequent users of istioctl, it can be useful to customize your default configuration, rather than typing it every time. We added the ability to put your personal defaults in your home directory. (Or wherever else you prefer.)
Human-readable text is easier than numbers β that’s why we have DNS! β so we also added it for port numbers. You can now specify port types using mnemonics like http instead of 80.
Unlike the Hotel California, you can both check out and leave, so we added ‘istioctl x uninstall’ to make that very easy.
Production operability improvements
Given Istio’s wide use in production systems, we’ve also made several improvements to its day 2 usability:
You can delay the application start until after the sidecar is started. This increases the reliability for deployments where the application needs to access resources via its proxy immediately upon its boot.
Sometimes stale endpoints could make Pilot become unhealthy. We fixed that.
The Istio Operator is a great way to install Istio, as it automates a fair amount of toil. Canary control plane deployments are also important; they allow ultra-safe upgrades of Istio. Unfortunately, you couldn’t use them together - until now.
We exposed metrics from the Istio-agent, so you can watch what’s going on with it.
We have made several improvements to our Prometheus metrics pipeline, getting more data there in an easier and more efficient manner.
VM support with added security
Since the early days of Istio, we’ve been working on support for incorporating workloads on VMs into a service mesh. While we’ve had users doing it for several releases now, with Istio 1.7 we have leaned in to add several improvements. Please note that this is still an Alpha feature.
One of the most used features of Istio is its security feature set. At its core is assigning a strong identity to each workload, in the form of short-lived certificates. In this release we are ensuring that workloads running on VMs in the mesh get a secure bootstrapping process, along with automatic certificate rotation.
For example, you might have a Kubernetes cluster hosting stateless web services (frontends) that serve data coming from stateful databases (backends) running in VMs outside of Kubernetes. You’d still like to encrypt the frontends’ accesses to these backends with mTLS. With this change, you can easily do that. Furthermore, this is done in a “zero trust” manner, where the compromise of one frontend or backend doesn’t allow the impersonation or compromise of the others, because the bootstrapping and certificate rotation is following best practices.
We also extended istioctl to be able to validate the proxy’s status for VM-based workloads, where validation was previously only available for Kubernetes-based workloads.
Finally we added official RPM packages, alongside the already-existing Debian packages. This should make installation on Red Hat-based images a very easy proposition.
Other fixes
We removed some invalid control plane metrics, and stopped installing telemetry addons by default.
We fixed an issue with SNI routing.
Istio now works better with headless services, as it will no longer send mTLS traffic to headless services without sidecars.
Join the Istio community
As always, there is a lot happening in the Community Meeting; join us every other Thursday at 10 AM Pacific. We’d love to have you join the conversation at Istio Discuss, and you can also join our Slack workspace.
We were very proud to be called out as one of the top five fastest growing open source projects in all of GitHub in 2019. Want to get involved? Join one of our Working Groups and help us make Istio even better.