Rucio Kubernetes Tutorial

Preliminaries

• Clone this repo to your local machine

• Install kubectl

  https://kubernetes.io/docs/tasks/tools/install-kubectl/
  

• Install helm

  https://helm.sh/docs/intro/install/
  

• Install minikube

  https://kubernetes.io/docs/tasks/tools/install-minikube/
  https://minikube.sigs.k8s.io/docs/start/
  

• Start minikube with extra RAM:

  minikube start --memory='4000mb'
  

• Add Helm chart repositories:

  helm repo add stable https://charts.helm.sh/stable
  helm repo add bitnami https://charts.bitnami.com/bitnami
  helm repo add rucio https://rucio.github.io/helm-charts
  

Some helpful commands

• Activate kubectl completion:

  # Bash
  source <(kubectl completion bash)
  # Zsh
  source <(kubectl completion zsh)
  

• View all containers:

  kubectl get pods / kubectl get pods --all-namespaces
  

• View/Tail logfiles of pod:

  kubectl logs <NAME>
  
  kubectl logs -f <NAME>
  

• Update helm repositories:

  helm repo update
  

• Shut down minikube:

  minikube stop
  

• For more useful commands:

1. kubectl : https://kubernetes.io/docs/reference/kubectl/cheatsheet/
2. helm : https://helm.sh/docs/helm/
3. minikube : https://cheatsheet.dennyzhang.com/cheatsheet-minikube-a4

Installation of Rucio + FTS + Storage

NOTE: Before executing the following commands, please change directory to the cloned repo location

NOTE: Replace the pod IDs with the ones from your instance, they change every time

NOTE: You can execute this Linux script for easier installation and understanding: ./automation/01-install-rucio.bash.

• Check if you have previously done this before and want to reset from scratch. In that case, check if there's an old PostgreSQL database lying around, and find and remove it with kubectl describe pvc && kubectl delete pvc data-postgres-postgresql-0

• If at any point helm fails to install, before re-installing, remove the previous failed installation by

  helm list # list all helm installations
  helm delete $installation

  You might also get errors that a job also exists. You can easily remove this by,

  kubectl get jobs # get all jobs
  kubectl delete jobs/$jobname

• Install a fresh new Rucio database (PostgreSQL).

  helm install postgres bitnami/postgresql -f values-postgres.yaml
  

• Wait for PostgreSQL to finish starting. Output should be STATUS:Running.

  kubectl get pods
  

  On bash/fish/zsh shells, you can also get the status of the job updated in real-time by

  (TOWATCH=postgres; watch "kubectl get pods --all-namespaces --no-headers |  awk '{if (\$2 ~ \"$TOWATCH\") print \$0}'")

  Here the fourth column represents the status.

  Tip: To watch the status any other install, just change value of TOWATCH from postgres to the name of your current install.

• Run init container once to setup the Rucio database once the PostgreSQL container is running:

  kubectl apply -f init-pod.yaml
  

• Watch the output of the init container to check if everything is fine. Pod should finish with STATUS:Completed

  kubectl logs -f init
  

  or on bash/fish/zsh shells by

  (TOWATCH=init; watch "kubectl get pods --all-namespaces --no-headers |  awk '{if (\$2 ~ \"$TOWATCH\") print \$0}'")

• Install the Rucio server and wait for it to come online:

  helm install server rucio/rucio-server -f values-server.yaml
  
  kubectl logs -f deployment/server-rucio-server rucio-server
  

• Prepare a client container for interactive use:

  kubectl apply -f client.yaml
  

• Once the client container is in STATUS:Running, you can jump into it and check if the clients are working:

  kubectl exec -it client /bin/bash
  
  rucio whoami
  

• Install the XRootD storage systems. This will start three instances of them.

  kubectl apply -f xrd.yaml
  

• Install the FTS database (MySQL) and wait for it to come online.

  kubectl apply -f ftsdb.yaml
  
  kubectl logs -f $(kubectl get pods -o NAME | grep fts-mysql | cut -d '/' -f 2)
  

For Windows:

  kubectl logs -f $(kubectl get pods -o name | findstr /c:"fts-mysql" | sed "s/^pod\///")

• Install FTS, once the FTS database container is up and running:

  kubectl apply -f fts.yaml
  
  kubectl logs -f $(kubectl get pods -o NAME | grep fts-server | cut -d '/' -f 2)
  

For Windows:

  kubectl logs -f $(kubectl get pods -o name | findstr /c:"fts-server" | sed "s/^pod\///")

• Install the Rucio daemons:

  helm install daemons rucio/rucio-daemons -f values-daemons.yaml
  

• Run FTS storage authentication delegation once:

  kubectl create job renew-manual-1 --from=cronjob/daemons-renew-fts-proxy
  

Rucio usage

NOTE: You can execute this Linux script for easier set up and understanding: ./automation/02-set-up-rucio.bash.

• Jump into the client container

  kubectl exec -it client /bin/bash
  

• Create the Rucio Storage Elements (RSEs) by

  rucio-admin rse add XRD1
  rucio-admin rse add XRD2
  rucio-admin rse add XRD3
  

• Add the protocol definitions for the storage servers

  rucio-admin rse add-protocol --hostname xrd1 --scheme root --prefix //rucio --port 1094 --impl rucio.rse.protocols.gfal.Default --domain-json '{"wan": {"read": 1, "write": 1, "delete": 1, "third_party_copy_read": 1, "third_party_copy_write": 1}, "lan": {"read": 1, "write": 1, "delete": 1}}' XRD1
  rucio-admin rse add-protocol --hostname xrd2 --scheme root --prefix //rucio --port 1094 --impl rucio.rse.protocols.gfal.Default --domain-json '{"wan": {"read": 1, "write": 1, "delete": 1, "third_party_copy_read": 1, "third_party_copy_write": 1}, "lan": {"read": 1, "write": 1, "delete": 1}}' XRD2
  rucio-admin rse add-protocol --hostname xrd3 --scheme root --prefix //rucio --port 1094 --impl rucio.rse.protocols.gfal.Default --domain-json '{"wan": {"read": 1, "write": 1, "delete": 1, "third_party_copy_read": 1, "third_party_copy_write": 1}, "lan": {"read": 1, "write": 1, "delete": 1}}' XRD3
  

• Enable FTS

  rucio-admin rse set-attribute --rse XRD1 --key fts --value https://fts:8446
  rucio-admin rse set-attribute --rse XRD2 --key fts --value https://fts:8446
  rucio-admin rse set-attribute --rse XRD3 --key fts --value https://fts:8446
  

  Note that 8446 is the port exposed by the fts-server pod. You can easily view ports opened by a pod by kubectl describe pod PODNAME.

• Fake a full mesh network

  rucio-admin rse add-distance --distance 1 --ranking 1 XRD1 XRD2
  rucio-admin rse add-distance --distance 1 --ranking 1 XRD1 XRD3
  rucio-admin rse add-distance --distance 1 --ranking 1 XRD2 XRD1
  rucio-admin rse add-distance --distance 1 --ranking 1 XRD2 XRD3
  rucio-admin rse add-distance --distance 1 --ranking 1 XRD3 XRD1
  rucio-admin rse add-distance --distance 1 --ranking 1 XRD3 XRD2
  

• Indefinite storage quota for root

  rucio-admin account set-limits root XRD1 -1
  rucio-admin account set-limits root XRD2 -1
  rucio-admin account set-limits root XRD3 -1
  

• Create a default scope for testing

  rucio-admin scope add --account root --scope test
  

• Create initial transfer testing data

  dd if=/dev/urandom of=file1 bs=10M count=1
  dd if=/dev/urandom of=file2 bs=10M count=1
  dd if=/dev/urandom of=file3 bs=10M count=1
  dd if=/dev/urandom of=file4 bs=10M count=1
  

• Upload the files

  rucio upload --rse XRD1 --scope test file1
  rucio upload --rse XRD1 --scope test file2
  rucio upload --rse XRD2 --scope test file3
  rucio upload --rse XRD2 --scope test file4
  

• Create a few datasets and containers

  rucio add-dataset test:dataset1
  rucio attach test:dataset1 test:file1 test:file2
  
  rucio add-dataset test:dataset2
  rucio attach test:dataset2 test:file3 test:file4
  
  rucio add-container test:container
  rucio attach test:container test:dataset1 test:dataset2
  

• Create a rule and remember returned rule ID

  rucio add-rule test:container 1 XRD3
  

• Query the status of the rule until it is completed. Note that the daemons are running with long sleep cycles (e.g. 30 seconds, 60 seconds) by default, so this will take a bit. You can always watch the output of the daemon containers to see what they are doing.

  rucio rule-info <rule_id>
  

  For this command, get the rule_id by,

  rucio list-rules test:container

• Add some more complications

  rucio add-dataset test:dataset3
  rucio attach test:dataset3 test:file4