github.com/grafana/tanka

Mini-release to fix an issue with our Makefile (required for packaging). No changes in functionality.

Bug Fixes

• build: Enable static Makefile target on all operating systems (#262)

Small patch release to fix a panic issue with tk apply.

Bug Fixes

• kubernetes: don't panic on failed diff (#256)

This release includes a critical fix, update ASAP.

Another Tanka release is here, just in time for Easter. Enjoy the built-in formatter, much more intelligent apply and several important bug fixes.

:rotating_light: Alert: kubectl diff changes resources :rotating_light:

The recently released kubectl version v1.18.0 includes a critical issue that causes kubectl diff (and so tk diff as well) to apply the changes.

This can be very harmful, so Tanka decided to require you to downgrade to v1.17.x, until the fix in kubectl version v1.18.1 is released.

• Upstream issue: https://github.com/kubernetes/kubernetes/issues/89762)
• Unreleased fix: https://github.com/kubernetes/kubernetes/pull/89795

:sparkles: Highlight: Jsonnet formatter (tk fmt)

Since jsonnetfmt was rewritten in Go recently, Tanka now ships it as tk fmt. Just run tk fmt . to keep all Jsonnet files recursively formatted.

:rocket: Highlight: Sorting during apply

When using tk apply, Tanka now automatically sorts your objects based on dependencies between them, so that for example CustomResourceDefinitions created before being used, all in the same run. No more partly failed applies!

Features

• kubernetes :sparkles:: Objects are now sorted by dependency before apply (#244)
• cli: Env var TANKA_KUBECTL_PATH can now be used to set a custom kubectl binary (#221)
• jsonnet :sparkles: : Bundle jsonnetfmt as tk fmt (#241)

• docker: The Docker image now includes GNU less, instead of the BusyBox one (#232)
• docker: Added kubectl, jsonnet-bundler, coreutils, git and diffutils to the Docker image, so Tanka can be fully used in there. (#243)

Bug Fixes

• cli: The diff shown on tk apply is now colored again (#216)

• client: The namespace patch file saved to a temporary location is now removed after run (#225)
• client: Scanning for the correct context won't panic anymore, but print a proper error (#228)
• client: Use os.PathListSeparator during context patching, so that Tanka also works on non-UNIX platforms (e.g. Windows) (#242)

• kubernetes :rotating_light:: Refuse to diff on kubectl version v1.18.0, because of above mentioned unfixed issue (#254)
• kubernetes: Apply no longer aborts when diff fails (#231)
• kubernetes :sparkles:: Namespaces that will be created in the same run are now properly handled during diff (#237)

Other

• cli: Migrates from spf13/cobra to much smaller go-clix/cli. This cuts our dependencies to a minimum. (#235)

(@xvzf) The next big one is here! Feature packed with environment overriding and tk export. Furthermore lots of bugs were fixed, so using Tanka should be much smoother now!

Highlight: Overriding vendor per Environment (#198)

It is now possible, to have a vendor/ directory managed by jb on an environment basis: https://tanka.dev/libraries/overriding. This means you can test out changes in libraries in single environments (like dev), without affecting others (like prod).

Notice:

Changes done in the last release (v0.7.1) can cause indentation changes when using std.manifestYAMLFromJSON(), related to bumping gopkg.in/yaml.v2 to gopkg.in/yaml.v3.
Please encourage all your teammembers to upgrade to at least v0.7.1 to avoid whitespace-only diffs on your projects.

Features

• cli: tk export can be used to write all generated Kubernetes resources to .yaml files

Bug Fixes

• kubernetes: Fail on diff when kubectl had an internal error (#213)
• kubernetes: Stop injecting namespaces into wrong places:
  Tanka was injecting the default namespace into resources of all kinds, regardless of whether they actually took one. This caused errors, so we stopped doing this. From now on, the default namespace will only be injected when the resource is actually namespaced. (#208)

• cli: tk diff colors:
  Before, the coloring was unstable when scrolling up and down. We fixed this by pressing CAPS-LOCK.
  Furthermore, the output of tk diff now also works on light color schemes, without messing up the background color. (#210)
• cli: Proper --version output:
  The release binaries now show the real semver on tk --version, instead of the git commit sha. (#201)
• cli: Print diff on apply again:
  While refactoring, we accidentally forgot to dereference a pointer, so that tk apply showed a memory address instead of the actual differences, which was kinda pointless. (#200)

This is a smaller release focused on critical bug fixes and some other minor enhancements. While features are included, none of them are significant, meaning they are part of a patch release.

Critical: parseYaml works now

Before, std.native('parseYaml') did not work at all, a line of code got lost during merge/rebase, resulting in parseYaml returning invalid data, that Jsonnet could not process. This issue has been fixed in (#195).

Jsonnet update

The built-in Jsonnet compiler has been upgraded to the lastest master 07fa4c0. In some cases, this should provide up to 50% more speed, especially when base64 is involved, which is now natively implemented. (#196)

Features

• cli: tk env set|add has been extended by --server-from-context, which allows to parse $KUBECONFIG to find the apiServer's IP directly from that file, instead of having to manually specify it by hand. (#184)
• jsonnet: vendor overrides:
  It is now possible to have a vendor/ directory per environment, so that updating upstream libraries can be done gradually. (#185)
• kubernetes: disable kubectl validation: tk apply now takes --validate=false to pass that exact flag to kubectl as well, for disabling the integrated schema validation. (#186)

Bug Fixes

• jsonnet, cli: Stable environment name: The value of (import "tk").env.name does not anymore depend on how Tanka was invoked, but will always be the relative path from <rootDir> to the environment's directory. (#182)
• jsonnet: The nativeFunc parseYaml has been fixed to actually return a valid result (#195)

The promised big update is here! In the last couple of weeks a lot has happened.

Grafana Labs announced Tanka to the public, and the project got a lot of positive feedback, shown both on HackerNews and in a 500+ increase in GitHub stars!

While we do not ship big new features this time, we ironed out many annoyances and made the overall experience a lot better:

Better website + tutorial (#134)

Our new website is published! It does not only look super sleek and performs like a supercar, we also revisited (and rewrote) the most of the content, to provide especially new users a good experience.

This especially includes the new tutorial, which gives new and probably even more experienced users a good insight into how Tanka is meant to be used.

:rotating_light::rotating_light: Disabling import ".yaml" (#176) :rotating_light::rotating_light:

Unfortunately, we had to disable the feature that allowed to directly import YAML files using the familiar import syntax, introduced in v0.6.0, because it caused serious issues with importstr, which became unusable.

While our extensions to the Jsonnet language are cool, it is a no-brainer that compatibility with upstream Jsonnet is more important. We will work with the maintainers of Jsonnet to find a solution to enable both, importstr and import ".yaml"

Workaround:

- import "foo.yaml"
+ std.parseYaml(importstr "foo.yaml")

k.libsonnet installation (#140)

Previously, installing k.libsonnet was no fun. While the library is required for nearly every Tanka project, it was not possible to install it properly using jb, manual work was required.

From now on, Tanka automatically takes care of this. A regular tk init installs everything you need. In case you prefer another solution, disable this new thing using tk init --k8s=false.

Features

• cli, kubernetes: k.libsonnet is now automatically installed on tk init (#140):
  Before, installing k.libsonnet was a time consuming manual task. Tanka now takes care of this, as long as jb is present on the $PATH. See https://tanka.dev/tutorial/k-lib#klibsonnet for more details.
• cli: tk env --server-from-context:
  This new flag allows to infer the cluster IP from an already set up kubectl context. No need to remember IP's anymore – and they are even autocompleted on the shell. (#145)
• cli, jsonnet: extCode, extVar:
  -e / --extCode and --extVar allow using std.extVar() in Tanka as well. In general, -e is the flag to use, because it correctly handles all Jsonnet types (string, int, bool). Strings need quoting! (#178)

• jsonnet: The contents of spec.json are now accessible from Jsonnet using (import "tk").env. (#163)
• jsonnet: Lists ([ ]) are now fully supported, at an arbitrary level of nesting! (#166)

Bug Fixes

• jsonnet: nil values are ignored from the output. This allows to disable objects using the if ... then {} pattern, which returns nil if false (#162).
• cli: -t / --target is now case-insensitive (#130)

---

First release of the new year! This one is a quick patch that lived on master for some time, fixing an issue with the recent "missing namespaces" enhancement leading to apply being impossible when no namespace is included in Jsonnet.

More to come soon :D

---

It has been quite some time since the last release during which Tanka has become much more mature, especially regarding the code quality and structure.

Furthermore, Tanka has just hit the 100 Stars :tada:

Notable changes include:

API (#97)

The most notable change is probably the Go API, available at https://godoc.org/github.com/grafana/tanka/pkg/tanka, which allows to use all features of Tanka directly from any other Golang application, without needing to exec the binary. The API is inspired by the command line parameters and should feel very similar.

Importing YAML (#106)

It is now possible to import .yaml documents directly from Jsonnet. Just use the familiar syntax import "foo.yaml" like you would with JSON.

Missing Namespaces (#120)

Tanka now handles namespaces that are not yet created, in a more user friendly way than `kubectl** does natively.
During diff, all objects of an in-existent namespace are shown as new and when applying, namespaces are applied first to allow applying in a single step.

Features

• tool/imports: import analysis using upstream jsonnet: Due to recent changes to google/jsonnet, we can now use the upstream compiler for static import analysis (#84)
• Array output: The output of Jsonnet may now be an array of Manifests. Nested arrays are not supported yet. (#112)

Bug Fixes

• Command Usage Guidelines: Tanka now uses the command description syntax (#94)
• cli/env resolved panic on missing spec.json (#108)

---

This version adds a set of commands to manipulate environments (tk env add, rm, set, list) (#73). The commands are mostly ks env compatible, allowing tk env be used as a drop-in replacement in scripts.

Furthermore, an error message has been improved, to make sure users can differentiate between parse issues in .jsonnet and spec.json (#71).

---

After nearly a month, the next feature packed release of Tanka is ready! Highlights include the new documentation website https://tanka.dev, regular expression support for targets, diff histograms and several bug-fixes.

Features

• cli: tk show now aborts by default, when invoked in a non-interactive session. Use --dangerous-allow-redirect to disable this safe-guard (#47).
• kubernetes: Regexp Targets: It is now possible to use regular expressions when specifying the targets using --target / -t. Use it to easily select multiple objects at once: https://tanka.dev/targets/#regular-expressions (#64).
• kubernetes: Diff histogram: Tanka now allows to summarize the differences between the live configuration and the local one, by using the unix diffstat(1) utility. Gain a sneek peek at a change using tk diff -s .! (#67)

Bug Fixes

• kubernetes: Tanka does not fail anymore, when the configuration file spec.json is missing from an Environment. While you cannot apply or diff, the show operation works totally fine (#56, #63).
• kubernetes: Errors from kubectl are now correctly passed to the user (#61).
• cli: tk diff does not output useless empty lines (\n) anymore (#62).

---

Tanka v0.3.0 is here!

This version includes lots of tiny fixes and detail improvements, to make it easier for everyone to configure their Kubernetes clusters.

Enjoy target support, enhancements to the diff UX and an improved CLI experience.

Features

The most important feature is target support (#30) (caf205a): Using --target=kind/name, you can limit your working set to a subset of the objects, e.g. to do a staged rollout.

There where some other features added:

• cli: autoApprove, forceApply (#35) (626b097): allows to skip the interactive verification. Furthermore, kubectl can now be invoked with --force.
• cli: print deprecated warnings in verbose mode. (#39) (6de170d): Warnings about the deprecated configs are only printed in verbose mode
• kubernetes: add namespace to apply preamble (#23) (9e2d927): The interactive verification now shows the metadata.namespace as well.
• cli: diff UX enhancements (#34) (7602a19): The user experience of the tk diff subcommand has been improved:
  • if the output is too long to fit on a single screen, the systems PAGER is invoked
  • if differences are found, the exit status is set to 16.
  • When tk apply is invoked, the diff is shown again, to make sure you apply what you want

Bug Fixes

• cli: invalid command being executed twice (#42) (28c6898): When the command failed, it was executed twice, due to an error in the error handling of the CLI.
• cli: config miss (#22) (32bc8a4): It was not possible to use the new configuration format, due to an error in the config parsing.
• cli: remove datetime from log (#24) (1e37b20)
• kubernetes: correct diff type on 1.13 (#31) (574f946): On kubernetes 1.13.0, subset was used, although native is already supported.
• kubernetes: Nil pointer deference in subset diff. (#36) (f53c2b5)
• kubernetes: sort during reconcile (#33) (ab9c43a): The output of the reconcilation phase is now stable in ordering

---