There are cases when the etcdserver is temporarily unavailable and the
errors that we get back from kube-apiserver reflect that error. It looks
like we bail out immediately when these errors happen currently. We
should retry until timeout is reached when this sort of errors happen.
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
Extend Interface with new InterfaceResources to avoid breaking changes
Move change to staus command behind --show-resources flag
Signed-off-by: Soujanya Mangipudi <somangip@microsoft.com>
Creating a new PR based on this existing stale PR https://github.com/helm/helm/pull/7728
Signed-off-by: Soujanya Mangipudi <somangip@microsoft.com>
# Conflicts:
# go.sum
Unfortunately errors from the API server do not always (do they ever?) contain
the name of the resource in question.
Deletions for multiple resources are processed concurrently, so in a resulting
log, a preceding "Starting delete" line might be for a different object.
Signed-off-by: Marcin Owsiany <porridge@redhat.com>
Confirm that the current and updated revision numbers also match as part
of the readiness check. Add coverage for readiness scenarios where
StatefulSet status does not reflect the most recent generation of the
StatefulSet yet.
Also add additional logging around the sts transitions from non-ready to
ready.
Fixes: #10163
Signed-off-by: Dominic Evans <dominic.evans@uk.ibm.com>
This required modifying the `kube.Factory` interface to conform to
changes in k8s' `cmdutil.Factory` interface:
fe3772890f
Signed-off-by: Andrew Seigner <andrew@sig.gy>
The partition value can be greater than number of replicas, in that
case no pods are rolled out. The expectedReplicas becomes a negative
number.
https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/#partitions
In the cases where the update does not change anything in the pod
template, the updatedReplicas value from StatefulSet status remains
unchanged. Such updates can still set some partition value, and
UpdatedReplicas is always greater than expectedReplicas. Basically,
the StatefulSet is ready / rolled-out.
In both the above scenarios, providing `--wait` flag causes it to
timeout waiting indefinitely. Because updatedReplicas can never be
negative, or be equal to the expectedReplicas for the second case.
This commit handles both the scenarios by checking if UpdatedReplicas
is smaller than expectedReplicas. If it is, then the StatefulSet is
not ready yet.
Based on the code from kubectl rollout:
a450ebd59c/pkg/polymorphichelpers/rollout_status.go (L138-L141)Closes#8674
Signed-off-by: Bhavin Gandhi <bhavin7392@gmail.com>
If set, 'uninstall' command will wait until all the resources are deleted before returning.
It will wait for as long as --timeout
closes#2378
Signed-off-by: Mike Ng <ming@redhat.com>
Fix typos
Remove condition arround time.Sleep
Because a negative or zero duration causes Sleep to return immediately.
Signed-off-by: Stephane Moser <moser.sts@gmail.com>
Rename the package time
Redesgin the logic to make a FakeKubeClient wait for a ammount time.
Remove unneed logic in the PrintingKubeClient
Signed-off-by: Stephane Moser <moser.sts@gmail.com>
Implement timer in the fake.go and printer.go to simulate the wait period
Add test Upgrade Release when it is interruped with SIGINT
Signed-off-by: Stephane Moser <moser.sts@gmail.com>
managedFields were a changed that landed in 1.18. This is an array
under metadata with managedFields. The kubernetes client pkgs that
Helm uses automatically add them.
This change added a manager for the managedFields. The flow for
deciding on the name to use is:
1. An explicit name if one is chosen
2. The base name of the first os.Arg (the binary name) if no name
explicitly set.
3. unknown if no name set and name cannot be detected
The name is at the package level as there is no other place to easily
set it for Helm v3. Since the name is for the binary or app it should
be ok to set app wide.
Signed-off-by: Matt Farina <matt.farina@suse.com>
Because backOffLimit can be 0, a zero value for pod status failed will
always cause the condition to return true.
Signed-off-by: James McElwain <jmcelwain@gmail.com>
Since Tiller is no longer part of Helm v3, internal documentation
language about Tiller can be removed
Signed-off-by: Matt Farina <matt@mattfarina.com>
* Continue deleting objects when one fails to minimize the risk of an
upgrade ending in an unrecoverable state
* Exclude failed deleted object from the returned result set
Signed-off-by: Adam Reese <adam@reese.io>
Upgrade Kubernetes libraries to v0.18.0
Add new lazy load KubernetesClientSet to avoid missing kubeconfig error
In kubernetes v1.18 kubeconfig validation was added. Minikube and Kind
both remove kubeconfig when stopping clusters. This causes and error
when running any helm commands because we initialize the client before
executing the command.
Signed-off-by: Adam Reese <adam@reese.io>
Don't delete a resource on upgrade if it is annotated with
helm.io/resource-policy=keep. This can cause data loss for users
if the annotation is ignored(e.g. for a PVC)
Close#7677
Signed-off-by: Dong Gang <dong.gang@daocloud.io>
But instead use a newly initialized Scheme with only Kubernetes native
resources added. This ensures the 3-way-merge patch strategy is not
accidentally chosen for custom resources due to them being added
to the global Scheme by e.g. versioned clients while using Helm as a
package, and not a self-contained binary.
Signed-off-by: Hidde Beydals <hello@hidde.co>
Fixes issue #7279.
Prevent the deletion of CRDs that were defined in the `templates/`
directory. This makes CRD deletion behaviour consistent with Helm
documentation:
> CRDs are never deleted. Deleting a CRD automatically deletes all of the
> CRD’s contents across all namespaces in the cluster. Consequently, Helm
> will not delete CRDs.
Previous the documentation only applied to CRDs that were defined in the
`crds/` directory. It did not consider that Charts could have CRDs in the
`templates/` directory (for example charts that were written before the
`crds/` directory feature or if the Chart author needed templated CRDs).
Signed-off-by: Phil Grayson <phil@philgrayson.com>
* Port watcher with retries to wait for resources
Port of Helm 2 PR #6014 to Helm 3
Signed-off-by: Martin Hickey <martin.hickey@ie.ibm.com>
* Add fix from PR #6907
Signed-off-by: Martin Hickey <martin.hickey@ie.ibm.com>
This was a missed update when we updated the k8s libraries. I validated
that this works for CRD installs with v1beta1 and v1
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
v3 port of #6792
After doing some more digging, I found out that updating the status
of an `Ingress` object is completely optional. Because of this, Helm
cannot support ingresses with the `--wait` flag because there is no
standard way to identify that they are ready
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
* Drop dependency on k8s.io/kubernetes
https://github.com/helm/helm/issues/6606
Depending on k8s.io/kubernetes is not recommended by Kubernetes, and
forces dependencies of Helm to also depend on them. We are only using
this dependency in one relatively isolated occurance, which can be
easily copied over rather than depending on the entire Kubernetes.
Copying this code is not very desirable, so if we don't want to have
this duplication we can at least use this PR as a PoC and see if we can
get Kubernetes to publish the controller package as a separate Go module
(see
https://github.com/kubernetes/kubernetes/issues/79384#issuecomment-538740756)
Signed-off-by: John Howard <howardjohn@google.com>
* Move to internal
Signed-off-by: John Howard <howardjohn@google.com>
* Exclude third_party from validate-license.sh
Signed-off-by: John Howard <howardjohn@google.com>
use *resource.Info.Namespace/Name because runtime.Object is a Interface
Signed-off-by: fengxusong <fengxsong@outlook.com>
Signed-off-by: Matthew Fisher <matt.fisher@microsoft.com>
Currently, if using the --atomic flag or deleting a release that failed due to an already existing
resource, Helm will deleting those resources that aren't managed by it. This PR fixes the issue
by checking for pre-existing resources during install and upgrade. This is done as a validation
step so the release will not even be started if resources currently exist. This PR is inspired by
@xchapter7x's work in #3477.
This also fixes a small bug in upgrade where deletes fail if the resource was already deletes
Fixes#6407
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
This ports the functionality of cleanup on fail to v3 as introduced in #4871. This has been tested manually
and would be a good candidate for a new acceptance test.
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
* fix: clear the discovery cache after CRDs are installed
This fixes an issue in which a chart could not contain both a CRD and an instance of that CRD. It works around a stale cache by force cache invalidation whenever a CRD is added.
Closes#6316
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
* fix: wait for CRD to register before allowing CRDs to be installed
This fixes an issue with the previous version of this patch in which the CRD would not be available quickly enough.
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
* feat: use Wait() to wait for CRDs to be ready
This forward-ports the CRD wait logic to Helm 3, and then uses that to wait for CRDs to be registered.
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
* ref: moved the scheme modification to an appropriate place.
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
* fix: turned warnings into fatal errors, fixed spelling, clear cache once
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
Sometimes the stateful set `rollingUpdate` field can be nil even when the
strategy is a rolling update
Fixes#6174
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
This changes most of the KubeClient interface to only ever build objects once and
then pass in everything as lists of resources. As a consequence, we needed to refactor
several of the actions. I took the opportunity to refactor out some duplicated
code while I was in the same area
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
`Result` is a misnomer and is going to be repurposed in a future commit for a
common result type for the different kube `Interface` methods
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
This fixes an issue where resources that hardcode the metadata.namespace parameter cannot be installed.
Signed-off-by: Matthew Fisher <matt.fisher@microsoft.com>
The partition and maxUnavailable values are now used in determining the state
of both objects
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
This reenables wait functionality and fixes some small bugs in the logic. Please
note that there are still some naive assumptions made about pods belonging to
DaemonSets and StatefulSets, but that is how the logic was before, so it was not
in scope to modify it for this PR. I will improve this logic in a follow up PR
Signed-off-by: Taylor Thomas <taylor.thomas@microsoft.com>
* move the main interface to it's own file
* removed summarizeKeptManifests() which was the last place kube.Get()
was called
* when polling for hooks, use external types
* refactor out legacyschema
* refactor detecting selectors from object
* refactor creating test client
Signed-off-by: Adam Reese <adam@reese.io>
This fixes a dozen or so style errors, almost all of which were just missing comments.
I left several which are fixed in other outstanding PRs, or which belong to code that is about to be removed.
Signed-off-by: Matt Butcher <matt.butcher@microsoft.com>
CRDs and other objects seen as unstructured cannot use strategic
merge patching. It has never been supported on CRDs. Previously,
cases like unstructured objects could have caused an unregistered
error. This is no longer the case.
This change explicitly looks for unstructured objects and handles
those using json merge patching.
Closes#3382
This error occures when resource is not found in helm release:
`Error: UPGRADE FAILED: no resource with the name "redis-cluster-sentinel" found`
Changed to:
`Error: UPGRADE FAILED: no ConfigMap with the name "redis-cluster-sentinel" found`
So now that resource can easily be found in cluster.
k8s client-go closes the ready channel that's passed in (see https://github.com/kubernetes/client-go/blob/master/tools/portforward/portforward.go#L171) This means that Tunnel's Close will always panic, as the client-go library will have closed then channel. This isn't reproducible unless helm.Client is externally, as the helm cli runner doesn't actually invoke Close.
Kind can change it's APIVersion with time, such as "Deployment" which was
"extensions/v1beta1" and then migrated to "apps/v1beta1" in Kube 1.6.
Name, Kind and optinally Namespace are used to find relevant Info
object for a original resource which is being upgraded, it safe enough,
as it is uniquely identifies an object.
This builds on previous work and adds more logging to follow
the full process of installing, updating, deleting, and rolling back.
This also standardizes capitalization of logs and small formatting
fixes
The current methodology generated its own RS slice instead of using
a helper method that uses a `clientset`. This caused some issues where
`FindNewReplicaSet` always returned `nil`. This switches the method and
removes some unneeded API calls and code.
Closes#2426
This fixes a bug in which 'helm status' fails if any of the expected
resources are missing in Kubernetes. Now it prints a list of missing
resources at the end of the status report.
Closes#2118
Headless services (with `clusterIP` set to `None`) were reporting
as not being ready because they didn't have an IP address. This adds
a logic check to ignore those services.
When manifests do not change, they didn't get populated with the
latest label info from the API server and then panic'd. This resolves
that bug
Closes#2043
* This is a simple mvp which processes a test definition with the
hook annotation for test when you run `helm test [release]`
* helm client cmd, proto def, tiller logic
Installs, rollback, upgrade, and delete now accept a `--timeout` flag
that allows the user to specify the maximum number of seconds that
any kubernetes command can take.
Closes#1678
It is unnecessary to create a namespace just to test that it exists.
Roles that don't have post namespace will error out even if the namespace exists
and it is possible to install the package anyway. This patch Gets a namespace and
only creates one if it doesn't exist.
Adds validation against the swagger schema.
Example error:
Error: release telling-wildebeest failed: error validating "": error
validating data: expected type int, for field
spec.template.spec.containers[0].ports[0].containerPort, got string
Current error:
unable to decode "": [pos 177]: json: expect char '"' but got char 'n'"'
This does the following to improve deletion failure handling:
- In an UninstallRelease operation, the release is marked DELETED
as soon as the basic checks are passed. This resolves 1508. I filed a
followup issue for doing this even better when we can modify protos
again.
- If a YAML manifest fails to parse, the error messages now suggests
that the record is corrupt, and the resources must be manually
deleted.
- If a resource is missing during deletion, the error messages now make
it clear that an object was skipped, but that the deletion continued.
Closes#1508
This feature has been disabled in the past because simply enabling the
feature causes the Kubernetes library to make requests to a server.
Thus, running any tests that use the 'pkg/kube' library has required
running a kube API server.
This patch makes it possible to selectively activate 3rdParty API
support, and then disables that support during testing.
Future versions of the Kubernetes library appear to make it easier to
configure and mock this behavior, so this is most likely a stop-gap
measure. (Famous last words.)
Closes#1472
- decouple tunnel from kube client
- add context switching for init cmd
- add unit tests for installer and init command
- refactor installer and remove unused code
Soujanya Mangipudi
Xin Ruan
Martin Hickey
Andrew Block
Qifan Shen
Cenk Alti
Davanum Srinivas
Marcin Owsiany
James Oden
Dominic Evans
Andrew Seigner
yxxhero
Matthew Fisher
Bhavin Gandhi
Jeff Rosenberg
Matt Farina
Mike Ng
lsowen
Stephane Moser
Rosenberg, Jeff
小龙同学
Matt Farina
Cookie Wang
Joe Lanford
James McElwain
Jon Huhn
zh168654
zhangye15
Janario Oliveira
Li Zhijian
Tariq Ibrahim
Rajat Jindal
Hidde Beydals
小明同学
Liu Ming
Adam Reese
Adam Reese
lnattrass
Dong Gang
Matthew Fisher
Taylor Thomas
ylvmw
Benn Linger
Federico Bevione
Jakub Bielecki
Ahmad Kazemi
Guangwen Feng
Patrik Cyvoct
Phil Grayson
Adrian Gonzalez-Martin
Vibhav Bobade
Fabian Ruff
John Howard
fengxsong
Charlie Getzen
Per Hermansson
Matt Butcher
Jacob LeGrone
Marc Khouzam
Michelle Noorali
AdamDang
Paolo Mainardi
Reinhard Nägele
Timofey Kirillov
Tim Hockin
Taylor Thomas
Matt Cholick
Kai Chen
Federico Gimenez
gardlt
Maxim Ivanov
devinyan
Justin Scott
peay
tyrannasaurusbanks
fengyun.pan
Kiichiro Okano
Kiichiro Okano
Matt Butcher
Anubhav Mishra
Matt Butcher
Taylor Thomas
Michelle Noorali
Vaughn Dice
Marius Grigoriu
Nikolay Mahotkin
Nikolay Mahotkin
Matt Butcher