“K8S Ecology Weekly” mainly contains some recommended weekly information related to K8S ecology that I have come into contact with. Welcome to subscribe zhihu column “K8S Ecology”.

Containerd V1.3.4 is released

Containerd v1.3.4 was released this week, about two months after containerd v1.3.3 was released. This is still the patch version of v1.3.x series, but there are a lot of notable content this time, let’s take a look:

  • Containerd /containerd#4007 improved the behavior of containerd host failed rollback in Docker Remote.

This change is very useful, for example, when we wanted to provide multiple mirror sources (or hosts) for a mirror repository. Prior to V1.3.4, if the first source (or host) returned an error, the default was to fall back and request the source directly. In this case, the equivalent of setting up multiple mirror sources (or hosts) is meaningless.

After this change, the request will try to set the mirror source one by one, which can greatly ensure the success rate of pull.

  • containerd/containerd#4104Added toNOTIFY_SOCKETSupport;
  • [Fixed] containerd/containerd#4150 [fixed] FIFO cleanup logic

This change is intended to fix a section of cleanup logic for FIFO that was introduced earlier. Prior to this change, if stdin/stdout/stderr was deleted using /dev/null, it would cause the /dev directory to be emptied when the task was closed, which would be very dangerous.

Refer to the release enote for additional changes to this release

Helm V2.16.6 and V3.2.0-RC.1 were released

This is a bugfix for Helm V2, which fixes a null pointer error #7812. If you still use Helm V2, you should upgrade to this version.

Helm V3 also released version V3.2.0-RC.1 this week. There are many notable changes in this release, but I will mention only one here, and the other changes will be discussed after the official release.

#7635 Added SQL storage driver support for Helm V3.

You can configure database connections using the HELM_DRIVER and HELM_DRIVER_SQL_CONNECTION_STRING variables.

V3.2 will be released soon, so you can check it out first.

Several versions of Kubernetes have been released

Kubernetes v1.16.9, v1.17.5 and v1.18.2 were released this week.

Notably for V1.18.2: #89908 fixes an issue that could cause Kubernetes crash after a node is deleted when deleting a Pod deployed on it triggers scheduling; (mainly because of its cache)

In general, it is very rare for a Node to be deleted before triggering the associated scheduling, and this problem may not be easy to encounter.

Progress in the upstream

  • # 89666 kubectl scaleArguments are now supported--dry-run=server|client!
  • # 89551To restore thekubectl apply --pruneCapability when namespace is not specified;

Recommended reading

  • Kubernetes. IO/blog / 2020/0… This section describes how to use Sidecar to monitor the services deployed in Kubernetes.

Please feel free to subscribe to my official account [MoeLove]