Upgrading

Upgrading Dotmesh on Docker

Download the latest stable dm client binary:

sudo curl -sSL -o /usr/local/bin/dm \
    https://get.dotmesh.io/$(uname -s)/dm
copy

Run:

dm cluster upgrade
copy

This will download the dotmesh-server docker image corresponding to the version of the dm client you’re using. It will then stop the Dotmesh server on the current node, and start the new version.

Run:

dm version
copy

To check that the version of the client and the server match, and are up-to-date.

Upgrading Dotmesh on Kubernetes

Download the latest stable dm client binary:

sudo curl -sSL -o /usr/local/bin/dm \
    https://get.dotmesh.io/$(uname -s)/dm
copy

Apply the latest stable dotmesh YAML for your Kubernetes version. For 1.7 generic:

kubectl apply -f https://get.dotmesh.io/yaml/configmap.yaml
kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-k8s-1.7.yaml
copy

For 1.7 on GKE:

kubectl apply -f https://get.dotmesh.io/yaml/configmap.gke.yaml
kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-k8s-1.7.yaml
copy

For 1.8 or 1.9 generic:

kubectl apply -f https://get.dotmesh.io/yaml/configmap.yaml
kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-k8s-1.8.yaml
copy

For 1.8 or 1.9 on GKE:

kubectl apply -f https://get.dotmesh.io/yaml/configmap.gke.yaml
kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-k8s-1.8.yaml
copy

For 1.8 on AKS:

kubectl apply -f https://get.dotmesh.io/yaml/configmap.aks.yaml
kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-k8s-1.8.aks.yaml
copy

You’ll notice the version of dotmesh-server is specified in the image tags within the Kubernetes YAML.

Run:

dm version
copy

To check that the version of the client and the server match, and are up-to-date.

dotmesh 0.4 -> 0.5

In versions 0.4 and below, Dotmesh was deployed via a DaemonSet, but version 0.5 introduces the Dotmesh Operator. Before upgrading to 0.5, please manually delete the Dotmesh DameonSet before running the normal upgrade procedure:

kubectl delete deployment dotmesh -n dotmesh
copy

dotmesh 0.3 -> 0.4

In the dotmesh 0.4 release, we upgraded the version of the etcd operator (0.5.0 -> 0.8.4) and etcd (3.1.8 -> 3.2.13). In order to be able to safely upgrade etcd and the etcd operator we added dm cluster backup-etcd and dm cluster restore-etcd commands. These commands let you tolerate the loss of an etcd cluster during an etcd upgrade.

It may also be useful to automate running the backup-etcd command regularly, so that you can recover from etcd data loss due to, for example, upgrading your GKE cluster or rebooting your nodes all at the same time – scenarios that the etcd operator deals with poorly. Future versions of dotmesh will automate continuous etcd backups.

So to upgrade dotmesh and the etcd operator and etcd on Kubernetes, follow these steps:

• Download the latest stable dm client binary (see here).

• Apply the appropriate dotmesh YAML for your Kubernetes type (see here).

• You will now have a dm client and a running cluster which supports backup-etcd and restore-etcd commands.

• Back up etcd (assuming your dm binary is pointing at the right cluster, use dm remote -v to check):

dm cluster backup-etcd > dotmesh-etcd-backup.json
copy

• Now, deploy the latest etcd operator:

kubectl apply -f https://get.dotmesh.io/yaml/etcd-operator-clusterrole.yaml
kubectl apply -f https://get.dotmesh.io/yaml/etcd-operator-dep.yaml
copy

• Delete your etcd cluster, it may well be in a failed state anyway (kubectl describe etcd dotmesh-etcd -n dotmesh if you are curious), and besides, we need to kick it to upgrade the etcd version itself:

kubectl delete etcd dotmesh-etcd-cluster -n dotmesh
copy

• Create a new etcd cluster:

kubectl apply -f https://get.dotmesh.io/yaml/dotmesh-etcd-cluster.yaml
copy

• Wait for the etcd cluster to come back up (look for three dotmesh-etcd-cluster-* pods to come up):

kubectl get pods -n dotmesh
copy

• Restore the backup.

dm cluster restore-etcd < dotmesh-etcd-backup.json
copy