Thomas Gummerer

When managing many resources with Pulumi, a pulumi up or pulumi destroy can lead to a complex graph of resource operations being performed as concurrently as possible relative to the dependencies in your program. If one of those operations fails, you may have noticed that Pulumi takes the most conservative approach, letting already in-flight operations complete, but not starting any new operations. Most often, that’s what you want – there might be no point in bringing up more infrastructure if a resource fails.

However, in some cases it can be useful to keep going to try to bring resources that are independent from the failed one into the requested state, be that resources being created or destroyed. For example, when doing a pulumi destroy, you may want to have Pulumi destroy as many resources as it can, without stopping when the first error occurs.

You can now do exactly that with the new --continue-on-error flag for pulumi up and pulumi destroy.

Read more →

At Pulumi, the stability of our CLI and SDK releases is critically important. A lot of infrastructure is built and deployed using Pulumi, and any bugs in how that infrastructure is deployed can cause outages. While we put a lot of emphasis on unit and integration testing, we are also constantly investing in new ways to provide additional safeguards to ensure we always deliver stable releases. To that end, we are introducing new dev releases of our CLI and SDKs, which will enable users who want to work with the very latest bits to try them out before they are deployed widely to the entire Pulumi user base. This blog post will walk you through how we are using them internally, and how you can too!

Read more →