Upgrade with Helm meet failure seems as configmap limitation exceed - kubernetes

I was using the helm upgrade -i xxx myfoo to install/upgrade myfoo,
I follow all the standard doc, but this failure ALWAYS happend as soon as I upgrade at the 7th upgration!
When I began to upgrade at the 7th time, it told me below failure:
Error: UPGRADE FAILED: ConfigMap "myfoo.v7" is invalid: []: Too long: must have at most 1048576 characters
This is really frustrating! Why do this happen?

Thank you for your help, I now understand what's going on.
I added some bigger file in chart/ dir, which cause the helm package size exceed 1M.
after remove these bigger files, it works back to normal. And I know why there is .helmignore file now, it is used to tell helm not include them in final helm package file(*.tgz)

Related

How to search for previous revisions using Microk8s etcd instance?

I'm trying to search previous versions of my secrets inside Microk8s etcd instance but the number of revisions change everytime I refresh my screen and I don't know why.
When I try to access an older version I get the error below:
etcdctl --endpoints=127.0.0.1:2380 get --rev=9133 -w fields /registry/secrets/default/mysql-test-password
{"level":"warn","ts":"2020-09-14T13:40:08.594Z","caller":"clientv3/retry_interceptor.go:61","msg":"retrying of unary invoker failed","target":"endpoint://client-8b3e59a8-efd2-4f77-a96f-5ec3c451b9b7/127.0.0.1:2380","attempt":0,"error":"rpc error: code = OutOfRange desc = etcdserver: mvcc: required revision has been compacted"}
Error: etcdserver: mvcc: required revision has been compacted
I also added the configuration on my etcd config file and restarted the service but didn't help:
--auto-compaction-mode=periodic
--auto-compaction-retention=72h
It seems that every time I refresh my screen the number of revisions increased a lot without doing anything.
{"header":{"cluster_id":14841639068965178418,"member_id":10276657743932975437,"revision":15322,"raft_term":7}
1 second later
student#desktop:~$ etcdctl --endpoints=127.0.0.1:2380 get /registry/secrets/default/mysql-root-password -w json
{"header":{"cluster_id":14841639068965178418,"member_id":10276657743932975437,"revision":16412,"raft_term":7}
Have someone faced something like that?
"etcdserver: mvcc: required revision has been compacted." is not an error, it's an expected message when a watch is attempted to be established at a resource version that has already been compacted.
Something is wrong with your etcd cluster, this is almost certainly not an apiserver problem. There's an alarm you might need to disable.
Also remember that you have to update etcd version to 3.0.11 or later -
https://github.com/kubernetes/kubernetes/issues/45506.
Take a look: mvcc-issues, reloader.

Kubernetes + Helm - only restart pods if new version/change

Whenever I run my basic deploy command, everything is redeployed in my environment. Is there any way to tell Helm to only apply things if there were changes made or is this just the way it works?
I'm running:
helm upgrade --atomic MyInstall . -f CustomEnvironmentData.yaml
I didn't see anything in the Helm Upgrade documentation that seemed to indicate this capability.
I don't want to bounce my whole evironment unless I have to.
There's no way to tell Helm to do this, but also no need. If you submit an object to the Kubernetes API server that exactly matches something that's already there, generally nothing will happen.
For example, say you have a Deployment object that specifies image: my/image:{{ .Values.tag }} and replicas: 3. You submit this once with tag: 20200904.01. Now you run the helm upgrade command you show, with that tag value unchanged in the CustomEnvironmentData.yaml file. This will in fact trigger the deployment controller inside Kubernetes. That sees that it wants 3 pods to exist with the image my/image:20200904.01. Those 3 pods already exist, so it does nothing.
(This is essentially the same as the "don't use the latest tag" advice: if you try to set image: my/image:latest, and redeploy your Deployment with this tag, since the Deployment spec is unchanged Kubernetes won't do anything, even if the version of the image in the registry has changed.)
You should probably use helm diff upgrade
https://github.com/databus23/helm-diff
$ helm diff upgrade - h
Show a diff explaining what a helm upgrade would change.
This fetches the currently deployed version of a release
and compares it to a chart plus values.
This can be used visualize what changes a helm upgrade will
perform.
Usage:
diff upgrade[flags] [RELEASE] [CHART]
Examples:
helm diff upgrade my-release stable / postgresql--values values.yaml
Flags:
-h, --help help for upgrade
--detailed - exitcode return a non - zero exit code when there are changes
--post - renderer string the path to an executable to be used for post rendering. If it exists in $PATH, the binary will be used, otherwise it will try to look for the executable at the given path
--reset - values reset the values to the ones built into the chart and merge in any new values
--reuse - values reuse the last release's values and merge in any new values
--set stringArray set values on the command line(can specify multiple or separate values with commas: key1 = val1, key2 = val2)
--suppress stringArray allows suppression of the values listed in the diff output
- q, --suppress - secrets suppress secrets in the output
- f, --values valueFiles specify values in a YAML file(can specify multiple)(default[])
--version string specify the exact chart version to use.If this is not specified, the latest version is used
Global Flags:
--no - color remove colors from the output

upgrade helm chart upon patching one of its objects

Do we need to explicitly upgrade helm chart "Release-1" when we do patch a particular object separately , for eg. Cron job "CJ1"?
In my case, I have patched cron job to run every min.
I did not however upgrade the helm chart that deployed the cron job.
"Kubectl get cj CJ1 -o yaml " although shows that the changes have been made from older schedule to the new schedule :- "* * * * *".
However the job is now not running at "* * * * *"
When you say patch I presume you are referring to editing the object with kubectl edit ... or in any other way that apply the change without going through helm upgrade ?
Generally speaking if you follow DevOps and GitOps best practices any change you do should go through git (be version controlled). If you patch an object separately/manually then your code no longer represent what you have deployed, so the next time you upgrade the chart you are going to get the version without the patch (loose your changes).
So if you want to keep the changes that you have applied separately/manually then Yes ... change your code, then upgrade the chart.
If in the long term it doesn't matter and you are just playing around ... then you don't have to do anything as the change you want is already in Kubernetes.

Security: Yaml Bomb: user can restart kube-api by sending configmap

Create yaml-bomb.yaml file:
apiVersion: v1
data:
a: &a ["web","web","web","web","web","web","web","web","web"]
b: &b [*a,*a,*a,*a,*a,*a,*a,*a,*a]
c: &c [*b,*b,*b,*b,*b,*b,*b,*b,*b]
d: &d [*c,*c,*c,*c,*c,*c,*c,*c,*c]
e: &e [*d,*d,*d,*d,*d,*d,*d,*d,*d]
f: &f [*e,*e,*e,*e,*e,*e,*e,*e,*e]
g: &g [*f,*f,*f,*f,*f,*f,*f,*f,*f]
h: &h [*g,*g,*g,*g,*g,*g,*g,*g,*g]
i: &i [*h,*h,*h,*h,*h,*h,*h,*h,*h]
kind: ConfigMap
metadata:
name: yaml-bomb
namespace: default
Send ConfigMap creation request to Kubernetes API by cmd kubectl apply -f yaml-bomb.yaml.
kube-api CPU/memory usage are very high, even later are getting restarted.
How do we prevent such yaml-bomb?
This is a billion laughts attack and can only be fixed in the YAML processor.
Note that the Wikipedia is wrong here when it says
A "Billion laughs" attack should exist for any file format that can contain references, for example this YAML bomb:
The problem is not that the file format contains references; it is the processor expanding them. This is against the spirit of the YAML spec which says that anchors are used for nodes that are actually referred to from multiple places. In the loaded data, anchors & aliases should become multiple references to the same object instead of the alias being expanded to a copy of the anchored node.
As an example, compare the behavior of the online PyYAML parser and the online NimYAML parser (full disclosure: my work) when you paste your code snippet. PyYAML won't respond because of the memory load from expanding aliases, while NimYAML doesn't expand the aliases and therefore responds quickly.
It's astonishing that Kubernetes suffers from this problem; I would have assumed since it's written in Go that they are able to properly handle references. You have to file a bug with them to get this fixed.
There's a couple of possible mitigations I could think of although as #flyx says the real fix here would be in the YAML parsing library used by Kubernetes.
Interestingly running this on a Kubernetes cluster on my local machine showed the CPU spike to be client-side (it's the kubectl process churning CPU) rather than server side.
If the issue was server side, then possible mitigations would be to use RBAC to minimize access to ConfigMap creation, and potentially to use an admission controller like OPA to review manifests before they are applied to the cluster.
This should probably be raised with the Kubernetes security vulnerability response team so that a proper fix can be implemented.
EDIT - I think where the problem manifests, might be down to the cluster version used. Server-side apply graduated to beta (should be enabled by default) in 1.16. So on a 1.16 cluster perhaps this would hit server side instead of client side.
EDIT - Just setup a 1.16 cluster, still showing the CPU usage as client-side in kubectl...
EDIT - I've filed an issue for this here also confirmed that the DoS can be achieved server-side by using curl instead of kubectl
Final EDIT - This got assigned a CVE (CVE-2019-11253) and is being fixed in Kubernetes 1.13+ . The fix has also been applied to the underlying YAML parsing lib here so any other Go programs should be ok as long as they're using an up to date version.
There was a TrustCom19 paper studying vulnerabilities in YAML parsers for different languages, it found that most parsers have some issues, so this is common and there are several recent CVEs in this space (details in paper: Laughter in the Wild: A Study into DoS Vulnerabilities in YAML Libraries, TrustCom19.
Preprint: https://www.researchgate.net/publication/333505459_Laughter_in_the_Wild_A_Study_into_DoS_Vulnerabilities_in_YAML_Libraries

Passing long configuration file to Kubernetes

I like the work methology of Kuberenetes, use self-contained image and pass the configuration in a ConfigMap, as a volume.
Now this worked great until I tried to do this thing with Liquibase container, The SQL is very long ~1.5K lines, and Kubernetes rejects it as too long.
Error from Kubernetes:
The ConfigMap "liquibase-test-content" is invalid: metadata.annotations: Too long: must have at most 262144 characters
I thought of passing the .sql files as a hostPath, but as I understand these hostPath's content is probably not going to be there
Is there any other way to pass configuration from the K8s directory to pods? Thanks.
The error you are seeing is not about the size of the actual ConfigMap contents, but about the size of the last-applied-configuration annotation that kubectl apply automatically creates on each apply. If you use kubectl create -f foo.yaml instead of kubectl apply -f foo.yaml, it should work.
Please note that in doing this you will lose the ability to use kubectl diff and do incremental updates (without replacing the whole object) with kubectl apply.
Since 1.18 you can use server-side apply to circumvent the problem.
kubectl apply --server-side=true -f foo.yml
where server-side=true runs the apply command on the server instead of the client.
This will properly show conflicts with other actors, including client-side apply and thus fail:
Apply failed with 4 conflicts: conflicts with "kubectl-client-side-apply" using apiextensions.k8s.io/v1:
- .status.conditions
- .status.storedVersions
- .status.acceptedNames.kind
- .status.acceptedNames.plural
Please review the fields above--they currently have other managers. Here
are the ways you can resolve this warning:
* If you intend to manage all of these fields, please re-run the apply
command with the `--force-conflicts` flag.
* If you do not intend to manage all of the fields, please edit your
manifest to remove references to the fields that should keep their
current managers.
* You may co-own fields by updating your manifest to match the existing
value; in this case, you'll become the manager if the other manager(s)
stop managing the field (remove it from their configuration).
See http://k8s.io/docs/reference/using-api/api-concepts/#conflicts
If the changes are intended you can simple use the first option:
kubectl apply --server-side=true -force-conflicts -f foo.yml
You can use an init container for this. Essentially, put the .sql files on GitHub or S3 or really any location you can read from and populate a directory with it. The semantics of the init container guarantee that the Liquibase container will only be launched after the config files have been downloaded.