-
Notifications
You must be signed in to change notification settings - Fork 0
Google Cloud: Deploying via Helm
Please Note: This is a UbiquityPress focused Wiki entry and does not apply to the Open Source Hyku Addons project
To build a new version of the application please refer to the HykuAddons: Building a new release for deployment article.
If everything is already configured and you're already connected to the correct cluster and region:
helm repo update;
helm upgrade hyku ubiquity-charts-hyku/hyku
https://helm.sh/docs/intro/install/
Start by using the helm list command to show the available releases.
~ ❯ helm list
NAME NAMESPACE REVISION UPDATED STATUS CHART APP VERSION
# ...
hyku default 30 2021-04-10 12:50:13.681628524 +0100 BST deployed hyku-4.0.1 2.1.0-194
# ...
If you see an error similar to the following, you may need to reauthorise GCloud:
Error: Kubernetes cluster unreachable: Get "https://xxx.xxx.xxx.xx/version?timeout=32s": error executing access token command "/snap/google-cloud-sdk/174/bin/gcloud config config-helper --format=json": err=fork/exec /snap/google-cloud-sdk/174/bin/gcloud: no such file or directory output= stderr=
Check that kubectl is authorised to read from the Hyku Addons cluster by using the kubectl get pods command.
If you see the following error, then you know you need to reauthorise your sesssion:
~ ❯ kubectl get pods
Unable to connect to the server: error executing access token command "/snap/google-cloud-sdk/174/bin/gcloud config config-helper --format=json": err=fork/exec /snap/google-cloud-sdk/174/bin/gcloud: no such file or directory output= stderr=
When kubectl works as exepected you should be able to use the helm list command and continue:
~ ❯ kubectl get pods
NAME READY STATUS RESTARTS AGE
# ...
hyku-66d7c77bdc-2b7sn 1/1 Running 0 10h
hyku-66d7c77bdc-8stzr 1/1 Running 0 10h
hyku-66d7c77bdc-pz68n 1/1 Running 0 45h
hyku-66d7c77bdc-r4mc6 1/1 Running 0 45h
# ...
~ ❯ helm list
NAME NAMESPACE REVISION UPDATED STATUS CHART APP VERSION
# ...
hyku default 30 2021-04-10 12:50:13.681628524 +0100 BST deployed hyku-4.0.1 2.1.0-194
# ...
Hyku Addons helm chart is stored in a seperate private repository, so much be added to your helm configuration manually before you can deploy.
First, search to ensure it has not yet been added:
~ ❯ helm search repo hyku
Error: no repositories configured
Like apt and other package managers you must add the repository before you can add the Helm chart.
~ ❯ helm repo add ubiquity-charts-hyku gs://helm-charts-hyku
"ubiquity-charts-hyku" has been added to your repositories
If you receive Error: could not find protocol handler for: gs, then you will need to add the GS protocol via a plugin for Helm.
~ ❯ helm plugin install https://github.com/hayorov/helm-gcs.git --version 0.3.9
Then you will need to reauthorise your GCloud account:
~ ❯ gcloud auth application-default login
Add the repository:
helm repo add ubiquity-charts-hyku gs://helm-charts-hyku
Now confirm that the repository was added successfully:
~ ❯ helm repo list
NAME URL
ubiquity-charts-hyku gs://helm-charts-hyku
Before you can deploy the application you will need to update the repository, much as you would with apt or brew.
~ ❯ helm repo update
Hang tight while we grab the latest from your chart repositories...
...Successfully got an update from the "ubiquity-charts-hyku" chart repository
Update Complete. ⎈Happy Helming!⎈
Now you should be able to search for the Hyku repository and see the current App version.
~ ❯ helm search repo hyku
NAME CHART VERSION APP VERSION DESCRIPTION
ubiquity-charts-hyku/hyku 4.0.1 2.1.0-194 Hyku repository
ubiquity-charts-hyku/tenant 0.0.3 2.5.68 Ubiquity Press repository tenant (with frontend)
Run the following to trigger the currently build version to be deployed:
~ ❯ helm upgrade hyku ubiquity-charts-hyku/hyku
Tenants have their own specialised import workers, to avoid a scenario whereby a large import negatively affects the performance of the user facing Hyku workers. If you have upgraded a cluster, you will also have to upgrade the import workers or recent changes to the codebase may not be available.
To check which release the node is currently using, you can use describe:
kubectl describe deployments demo-import-workers | grep Image
# => Image: eu.gcr.io/hyku-ucw/hyku:2.1.0-400Normally, to update the Hyku node, you use the following:
# Update the chart
helm repo update; helm upgrade hyku ubiquity-charts-hyku/hykuThe import workers however, are not upgraded with this command. This is because the hyku Helm chart contains only the "hyku" and "workers" deployments, so the helm upgrade hyku #... command, specifies that only "hyku" should be updated.
To update a tenants import workers, which are labelled "<tenant_name>-import-workers", to the latest release:
# helm upgrade <tenant_name> ubiquity-charts-hyku/tenant
# So, to update the "demo" tenant
helm upgrade demo ubiquity-charts-hyku/tenantTo confirm it worked, run the describe command again:
kubectl describe deployments demo-import-workers | grep Image
# => Image: eu.gcr.io/hyku-ucw/hyku:2.1.0-401Please note: this will also upgrade the frontend codebase, so you should confirm that this is ok with Rees before proceeding.
If you wish to upgrade all of the import workers, you will have to manually run the upgrade command for each tenant. This means you need to know the name of each of the tenants in order to pass it to the upgrade command. This can be done by calling kubectl describe deployments without passing anything else. You will see a list of details for each of the nodes, including the import workers, and other services that you dont need to worry about.
In order to make it a little easier to parse, you can run the following command:
kubectl describe deployments | grep ^Name: | grep -v [-] | awk '{ print $NF }'This will look for just the name of the node and ignore any with hyphenated names like "demo-import-workers".
anschutz
anschutzdemo
edemo
fcrepo
hyku
ltuscdemo
nsuoklib
nsuoklibdemo
pacific
pacificdemo
redlands
redlandsdemo
una
unademo
workersAs you can see from this list, there are a number that you do not need to worry about, like "fcrepo", "workers", "hyku". You will have to parse out the ones you are not interested from the list.
You can use helm get notes hyku to get the default super admin password for the cluster you are connected to.
To list all our deployments for the cluster where your images are stored - we are using the US cluster as an example:
gcloud config configurations activate ubiquityrepo-us;
gcloud container clusters get-credentials cluster-us --zone us-central1-a
Run the command to show current deployments:
kubectl get deployments
This should return you a table of current deployment:
NAME READY UP-TO-DATE AVAILABLE AGE
fcrepo 1/1 1 1 119d
hyku 4/4 4 4 119d
mailgun-operator 1/1 1 1 120d
pacific 2/2 2 2 86d
pacificdemo 2/2 2 2 118d
workers 2/2 2 2 119d
In this list you can see hyku and workers. These two belong to the Hyku Helm chart and will be updated when you run helm upgrade hyku. However, this list also contains some specialised tenant workers, such as anschutzdemo-import-workers.
Get the current build version for the container:
kubectl describe deployments hyku | grep ImageTo see which build versions each of these pods is current running:
kubectl describe deployments workers
You will see a complete description of the deployment, containing the Image version, similar to this:
Image: eu.gcr.io/hyku-ucw/hyku:2.1.0-288
However, if you look at the specialised tenant import worker pods:
kubectl describe deployments anschutzdemo-import-workers
You may well see a different image version in use:
Image: eu.gcr.io/hyku-ucw/hyku:2.1.0-285