Etcd is a highly available distributed key value store that provides a reliable way to store data across a cluster of machines. Etcd gracefully handles master elections during network partitions and will tolerate machine failure, including the master.

Your applications can read and write data into etcd. A simple use-case is to store database connection details or feature flags in etcd as key value pairs. These values can be watched, allowing your app to reconfigure itself when they change.

Advanced uses take advantage of the consistency guarantees to implement database master elections or do distributed locking across a cluster of workers.

Etcd allows storing data in a distributed hierarchical database with observation.

We can deploy a single node with the following commands:

juju deploy easyrsa
juju deploy etcd
juju add-relation etcd easyrsa

And add capacity with:

juju add-unit -n 2 etcd

It's recommended to run an odd number of machines as it has greater redundancy than an even number (i.e. with 4, you can lose 1 before quorum is lost, whereas with 5, you can lose 2).

The etcd charm initializes a cluster using the Static configuration: which is the most "flexible" of all the installation options, considering it allows etcd to be self-discovering using the peering relationships provided by Juju.

Health of the cluster can be checked by running a juju action.

juju action do etcd/0 health
<return response uuid>
juju action fetch <uuid>

The health is also reported continuously via juju status. During initial cluster turn-up, it's entirely reasonable for the health checks to fail; this is not a situation to cause you alarm. The health-checks are being executed before the cluster has stabilized, and it should even out once the members start to come online and the update-status hook is run again.

This will give you some insight into the cluster on a 5 minute interval, and will report healthy nodes vs unhealthy nodes.

For example:

Unit        Workload  Agent  Machine  Public address  Ports     Message
etcd/0*     active    idle   1        54.227.0.225    2379/tcp  Healthy with 3 known peers
etcd/1      active    idle   2        184.72.191.212  2379/tcp  Healthy with 3 known peers
etcd/2      active    idle   3        34.207.195.139  2379/tcp  Healthy with 3 known peers

The ETCD charm supports TLS terminated endpoints by default. All efforts have been made to ensure the PKI is as robust as possible.

Client certificates can be obtained by running an action on any of the cluster members:

juju run-action etcd/0 package-client-credentials
juju scp etcd/0:etcd_credentials.tar.gz etcd_credentials.tar.gz

This will place the client certificates in pwd. If you're keen on using etcdctl outside of the cluster machines, you'll need to expose the charm, and export some environment variables to consume the client credentials.

juju expose etcd
export ETCDCTL_KEY_FILE=$(pwd)/client.key
export ETCDCTL_CERT_FILE=$(pwd)/client.crt
export ETCDCTL_CA_FILE=$(pwd)/ca.crt
export ETCDCTL_ENDPOINT=https://{ip of etcd host}:2379
etcdctl member list

Many cloud providers use ephemeral storage. When using cloud provider infrastructures is recommended to place any data-stores on persistent volumes that exist outside of the ephemeral storage on the unit.

Juju abstracts this with the storage provider.

To add a unit of storage we'll first need to discover what storage types the cloud provides to us, which can be listed:

juju list-storage-pools

To add SSD backed EBS storage from AWS, the following example provisions a single 10GB SSD EBS instance and attaches it to the etcd/0 unit.

juju add-storage etcd/0 data=ebs-ssd,10G

To add Persistent Disk storage from GCE, the following example provisions a single 10GB PD instance and attaches it to the etcd/0 unit.

juju add-storage etcd/0 data=gce,10G

To add Persistent Disk storage from Open Stack Cinder, the following example provisions a single 10GB PD instance and attaches it to the etcd/0 unit.

juju add-storage etcd/0 data=cinder,10G

Allows the operator to restore the data from a cluster-data snapshot. This comes with caveats and a very specific path to restore a cluster:

The cluster must be in a state of only having a single member. So it's best to deploy a new cluster using the etcd charm, without adding any additional units. Snapshots are also only supported on the same etcd version with which the snapshot was taken.

juju deploy etcd new-etcd

The above code snippet will deploy a single unit of etcd, as 'new-etcd' Don't forget to juju attach the snapshot created in the snapshot action.

juju attach new-etcd snapshot=/path/to/etcd-backup
juju run-action new-etcd/0 restore

Once the restore action has completed, evaluate the cluster health. If the cluster is healthy, you may resume scaling the application to meet your needs.

• param target: destination directory to save the existing data.

• param skip-backup: Don't backup any existing data. (defaults to True)

Allows the operator to snapshot a running clusters data for use in cloning, backing up, or migrating etcd clusters.

juju run-action etcd/0 snapshot target=/mnt/etcd-backups

• param target: destination directory to save the resulting snapshot archive.

Migrating the etcd data is a fairly easy task. Use the following steps:

Step 1: Snapshot your existing cluster. This is encapsulated in the snapshot action.

$ juju run-action etcd/0 snapshot

Action queued with id: b46d5d6f-5625-4320-8cda-b611c6ae580c

Step 2: Check the status of the action so you can verify the hash sum of the resulting file. The output will contain results.copy.cmd the value can be copied and used to download the snapshot that you just created.

Download the snapshot tar archive from the unit that created the snapshot and verify the sha256 hash sum.

$ juju show-action-output b46d5d6f-5625-4320-8cda-b611c6ae580c
results:
  copy:
    cmd: juju scp etcd/0:/home/ubuntu/etcd-snapshots/etcd-snapshot-2016-11-09-02.41.47.tar.gz
      .
  snapshot:
    path: /home/ubuntu/etcd-snapshots/etcd-snapshot-2016-11-09-02.41.47.tar.gz
    sha256: 1dea04627812397c51ee87e313433f3102f617a9cab1d1b79698323f6459953d
    size: 68K
status: completed

$ juju scp etcd/0:/home/ubuntu/etcd-snapshots/etcd-snapshot-2016-11-09-02.41.47.tar.gz .

$ sha256sum etcd-snapshot-2016-11-09-02.41.47.tar.gz

Step 3: Deploy the new cluster leader, and attach the snapshot as a resource.

juju deploy etcd new-etcd --resource snapshot=./etcd-snapshot-2016-11-09-02.41.47.tar.gz

Step 4: Re-Initialize the etcd leader with the data by running the restore action which uses the resource that was attached in step 3.

juju run-action new-etcd/0 restore

Step 5: Scale and operate as required, verify the data was restored.

The etcd charm installs the etcd application as a snap package. You can supply an etcd.snap resource to make this charm easily installable behind a firewall.

juju deploy /path/to/etcd
juju attach etcd etcd=/path/to/etcd.snap

The charm if installed from a locally supplied resource will be locked into that resource version until another is supplied and explicitly installed.

juju attach etcd etcd=/path/to/new/etcd.snap
juju run-action etcd/0 install
juju run-action etcd/1 install

This section only applies if you are upgrading an existing etcd charm deployments. This migration should only be needed once because new deployments of etcd will default to snap delivery.

Revision 24 and prior the etcd charm installed the etcd application from Debian packages. Revisions 25+ install from the snap store (or resource). During the migration process, you will be notified that a classic installation exists and a manual migration action must be run.

Before a migration is your opportunity to ensure state has been captured, and to plan for downtime, as this migration process will stop and resume the etcd application. This service disruption can cause disruptions with other dependent applications.

The deb to snap migration process has been as automated as possible. Despite the automatic backup mechanism during the migration process, you are still encouraged to run a snapshot before executing the upgrade.

Once the snapshot is completed, begin the migration process. You first need to upgrade the charm to revision 25 or later.

juju upgrade-charm etcd

For your convenience there is the snap-upgrade action that removes the Debian package and installs the snap package. Each etcd unit will need to be upgraded individually. Best practice would be to migrate an individual unit at a time to ensure the cluster upgrades completely.

juju run-action etcd/0 snap-upgrade  
# Repeat this command for other etcd units in your cluster.

Once the unit has completed upgrade, the unit's status message will return to its normal health check messaging.

Unit        Workload  Agent  Machine  Public address  Ports     Message
etcd/0*     active    idle   1        54.89.190.93    2379/tcp  Healthy with 3 known peers

Once you have the snap package you can upgrade to different versions of etcd by configuring the snap channel configuration option on the charm.

juju config etcd channel=3.0/stable

The etcd charm relies heavily on the snap package for etcd. In order to properly migrate a 2.x series etcd deployment into 3.1 and beyond you will need to follow a proper channel migration path. The initial deb to snap upgrade process will place you in a 2.3 deployment.

You can migrate from 2.3 to 3.0

juju config etcd channel=3.0/stable

From the 3.0 channel you can migrate to 3.1 (current latest at time of writing)

juju config etcd channel=3.1/stable

You MUST perform the 2.3 => 3.0 before moving from 3.0 => 3.1 A migration from 2.3 => 3.1 is not supported at this time.

Additionally, this charm breaks with no backwards compatible/upgrade path at the Trusty/Xenial series boundary. Xenial forward will enable TLS by default. This is an incompatible break due to the nature of peer relationships, and how the certificates are generated/passed off.

To migrate from Trusty to Xenial, the operator will be responsible for deploying the Xenial etcd cluster, then issuing an etcd data dump on the trusty series, and importing that data into the new cluster. This can be only be performed on a single node due to the nature of how replicas work in etcd.

Any issues with the above process should be filed against the charm layer in github.

Restoring from a snapshot on a scaled cluster will result in a broken cluster. Etcd performs clustering during unit turn-up, and state is stored in etcd itself. During the snapshot restore phase, a new cluster ID is initialized, and peers are dropped from the snapshot state to enable snapshot restoration. Please follow the migration instructions above in the restore action description.

• Charles Butler <[email protected]>
• Mathew Bruzek <[email protected]>