Charmed MongoDB

Charmed MongoDB Operator

Description

The Charmed MongoDB Operator delivers automated operations management from day 0 to day 2 on the MongoDB Community Edition document database.

MongoDB is a general purpose distributed document database. This operator charm deploys and operates MongoDB (vs 5.0) on physical or virtual machines as a single replica.

Overview

This charm deploys MongoDB in three configurations:

• Single node
• Replica set
• Sharded clusters

By default, the MongoDB application is installed from the Ubuntu archive, except for arm64 platforms. The version of MongoDB in the archive is known to have issues on arm64, so by default this charm will use ppa:mongodb-arm64/ppa which contains backported fixes for this architecture.

Usage

Review the configurable options

The MongoDB charm allows for certain values to be configurable via a config.yaml file. The options provided are extensive, you should review the options.

Specifically the following options are important:

• replicaset

  • ie: myreplicaset
  • Each replicaset has a unique name to distinguish it’s members from other replicasets available in the network.
  • The default value of "myset" should be fine for most single cluster scenarios.

• web_admin_ui

  • MongoDB comes with a basic but very informative web user interface that provides health and status information on the database node as well as the cluster.
  • The default value of yes will start the Admin web UI on port 28017.

Most of the options in config.yaml have been modeled after the default configuration file for mongodb (normally in /etc/mongodb.conf) and should be familiar to most mongodb admins. Each option in this charm have a brief description of what it does.

Usage

Single Node

Deploy the first MongoDB instance

juju deploy mongodb
juju expose mongodb

Replica Sets

Deploying

Deploy the first two MongoDB instances that will form replicaset:

juju deploy mongodb -n 2

Deploying three or more units at start can sometimes lead to unexpected race-conditions so it's best to start with two nodes.

Your deployment should look similar to this ( juju status ):

environment: amazon
machines:
  "0":
    agent-state: started
    agent-version: 1.16.5
    dns-name: ec2-184-73-7-172.compute-1.amazonaws.com
    instance-id: i-cb55cceb
    instance-state: running
    series: precise
    hardware: arch=amd64 cpu-cores=1 cpu-power=100 mem=1740M root-disk=8192M
  "1":
    agent-state: pending
    dns-name: ec2-54-196-181-161.compute-1.amazonaws.com
    instance-id: i-974bd2b7
    instance-state: pending
    series: precise
    hardware: arch=amd64 cpu-cores=1 cpu-power=100 mem=1740M root-disk=8192M
services:
  mongodb:
    charm: cs:precise/mongodb-20
    exposed: false
    relations:
      replica-set:
      - mongodb
    units:
      mongodb/0:
        agent-state: pending
        machine: "1"
        public-address: ec2-54-196-181-161.compute-1.amazonaws.com

In addition, the MongoDB web interface should also be accessible via the services’ public-address and port 28017 ( ie: http://ec2-50-17-73-255.compute-1.amazonaws.com:28017 ).

(Optional) Change the replicaset name

juju config mongodb replicaset=<new_replicaset_name>

Add one or more nodes to your replicaset

juju add-unit mongodb
juju add-unit mongodb -n2

We now have a working MongoDB replica-set.

Caveats

Keep in mind that you need to have odd number of nodes for a properly formed replicaset.

Replicaset can't function with only one available node - shall this happens the remaining node is switched to 'read-only' until at least one of the broken nodes is restored.

More info can be found in MongoDB documentation at their website

Removing a failed node

Working units can be removed from replica set using 'juju remove-unit' command. If the removing unit is primary it will automatically be stepped down (so thath re-election of new primary is performend) before being removed. However, if a unit fails (freezes, gets destroyed and is unbootable), operator needs to manually remove it. The operator would connect to primary unit, and issue rs.remove() for failed unit. Also, operator needs to issue 'juju remove-unit --force' to remove failed unit from juju.

Recovering from degraded replicaset

If two members go down replicaset is in read-only state. That is because the remaining node is in SECONDARY state (it can't get promoted/voted to PRIMARY because there is no majority in replicaset). If failed nodes can't be brought back to life we need to manually force remaining node to become a primary. Here is how:

1. connect to the node that's alive

2. start 'mongo', a cli utility

3. upon connecting you'll see that node is SECONDARY

4. display current configuration with: rs.config()

   • this will show the alive node as well as the nodes that are unreachabble

5. store the configuration into some temporary json document: cfg=rs.config()

6. change the cfg document so that it's members array contain only the unit that is alive: cfg.members=[cfg.members[0]]

7. force reconfiguration of the replicaset: rs.reconfig(cfg, {force: true})

8. wait a few, and press ENTER. You should see that your node becomes PRIMARY.

After this clean up the unavailable machines from juju: juju remove-machine --force XX ## XX is the machine number

And add more units to form a proper replicaset. (To avoid race conditions it is best to add units one by one).

   juju add-unit mongodb

Enable authentication

To enable authentication the property auth should be set.

juju config mongodb auth=true

When authentication is enabled a user is created so that the charm can communicate with the mongodb server. The credentials for this user are stored in the file /var/lib/mongodb/charm_pass.conf. This file can only be read by root.

Enabling authentication breaks the connection from related database client units as the password is not passed through the relation.

Sharding (Scale Out Usage)

According the the MongoDB documentation found on their website, one way of deploying a Shard Cluster is as follows:

• deploy config servers
• deploy a mongo shell (mongos)
• deploy shards
• connect the config servers to the mongo shell
• add the shards to the mongo shell

Using Juju we can deploy a sharded cluster using the following commands:

Prepare a configuration file similar to the following:

shard1:
  replicaset: shard1
shard2:
  replicaset: shard2
shard3:
  replicaset: shard3
configsvr:
  replicaset: configsvr

We'll save this one as ~/mongodb-shard.yaml.

Bootstrap the environment

juju bootstrap

Config Servers ( we'll deploy 3 of them )

juju deploy mongodb configsvr --config ~/mongodb-shard.yaml -n3

Mongo Shell ( We just deploy one for now )

juju deploy mongodb mongos

Shards ( We'll deploy three replica-sets )

juju deploy mongodb shard1 --config ~/mongodb-shard.yaml -n3
juju deploy mongodb shard2 --config ~/mongodb-shard.yaml -n3
juju deploy mongodb shard3 --config ~/mongodb-shard.yaml -n3

Connect the Config Servers to the Mongo shell (mongos)

juju add-relation mongos:mongos-cfg configsvr:configsvr

Connect each Shard to the Mongo shell (mongos)

juju add-relation mongos:mongos shard1:database
juju add-relation mongos:mongos shard2:database
juju add-relation mongos:mongos shard3:database

With the above commands, we should now have a three replica-set sharded cluster running. Using the default configuration, here are some details of our sharded cluster:

• mongos is running on port 27021
• configsvr is running on port 27019
• the shards are running on the default mongodb port of 27017
• The web admin is turned on by default and accessible with your browser on port 28017 on each of the shards.

To verify that your sharded cluster is running, connect to the mongo shell and run sh.status():

• mongo --host <mongos_host>:<mongos_port>
• run sh.status()

You should see your the hosts for your shards in the status output.

Use the storage subordinate to store mongodb data on a permanent OpenStack or Amazon EBS volume

The storage subordinate and block-storage-broker service can automatically handle attaching the volume and mounting it to the unit before MongoDB is setup to use it.

For example if you've created the volumes vol-id-00001 and vol-id-00002 and want to attach them to your 2 mongo units, with your OpenStack or AWS credentials in a credential.yaml file:

juju deploy block-storage-broker --config credentials.yaml
juju deploy storage
juju add-relation block-storage-broker storage
juju config storage provider=block-storage-broker
juju config volume_map="{mongodb/0: vol-id-00001, mongodb/1: vol-id-00002}"
juju add-relation storage mongodb

Use a permanent Openstack volume to store mongodb data. (DEPRECATED)

Note: Although these steps will still work they are now deprecated, you should use the storage subordinate above instead.

To deploy mongodb using permanent volume on Openstack, the permanent volume should be attached to the mongodb unit just after the deployment, then the configuration should be updated like follows.

juju config mongodb volume-dev-regexp="/dev/vdc" volume-map='{"mongodb/0": "vol-id-00000000000000"}' volume-ephemeral-storage=false

Backups

Backups can be enabled via config. Note that destroying the service cannot currently remove the backup cron job so it will continue to run. There is a setting for the number of backups to keep, however, to prevent from filling disk space.

To fetch the backups scp the files down from the path in the config.

Use Percona MongoDB

You can enable the Percona Mongodb server by configuring the source parameter with the percona repository. You can use the following as an example:

# config.yaml content
mongodb:
  package: percona-server-mongodb
  source: http://repo.percona.com/psmdb-44/apt  main
  key: 9334A25F8507EFA5

juju deploy --config config.yaml  cs:mongodb

Known Limitations and Issues

• If your master/slave/replicaset deployment is not updating correctly, check the log files at /var/log/mongodb/mongodb.log to see if there is an obvious reason ( port not open etc.).
• Ensure that TCP port 27017 is accessible from all of the nodes in the deployment.
• If you are trying to access your MongoDB instance from outside your deployment, ensure that the service has been exposed ( juju expose mongodb )
• Make sure that the mongod process is running ( ps -ef | grep mongo ).
• Try restarting the database ( restart mongodb )
• If all else fails, remove the data directory on the slave ( rm -fr /var/log/mongodb/data/* ) and restart the mongodb-slave daemon ( restart mongodb ).
• Configuration options "cpu", "quota", "rest", "noprealloc", "nssize", "master", "arbiter", "autoresync", "extra_config_options" are deprecated >= 4.2 (https://docs.mongodb.com/manual/reference/configuration-options/#removed-mmapv1-options)

Contact Information

MongoDB Contact Information

• MongoDB website
• MongoDB documentation
• MongoDB bug tracker
• MongoDB user mailing list

License

The Charmed MongoDB Operator is distributed under the Apache Software License, version 2.0. It [installs/operates/depends on] MongoDB Community Edition, which is licensed under the Server Side Public License (SSPL)

Trademark notice

MongoDB' is a trademark or registered trademark of MongoDB Inc. Other trademarks are property of their respective owners.