Configuring a MongoDB Sharded Cluster with PMM2 – Part 2

Configure MongoDB Sharded Cluster As a DBA, it is important to monitor a database to help us troubleshoot or to understand the health of an instance. Percona Monitoring and Management (PMM v2) is open-source and does a great job in monitoring the databases like MongoDB, MySQL, PostgreSQL, etc.

In this blog post, we will see how to configure a sharded cluster in PMM2. This is a part two version of the previous one which was done with PMM v1, titled Configuring PMM Monitoring for MongoDB Cluster. I have listed the steps to configure the sharded cluster into PMM2 below:

Prepare DB for Monitoring

Before configuring with PMM2, we will need to create a USER for monitoring from the database side. If you need to enable QAN (query analytics), then you will need to enable profiler and some more custom permission like “explainRole” to the user as well. Adding profiler adds up some more little load to the database, so it is better you do prior tests to analyze the load if you want to assess the extra load.

Add PMM Users to the DB

user

Shell

// Change role name / user / password as required

db.getSiblingDB("admin").createRole({

role: "explainRole",

privileges: [{

resource: {

db: "",

collection: ""

actions: [

"listIndexes",

"listCollections",

"dbStats",

"dbHash",

"collStats",

"find"

roles:[]

db.getSiblingDB("admin").createUser({

user: "pmm_mongodb",

pwd: "password",

roles: [

{ role: "explainRole", db: "admin" },

{ role: "clusterMonitor", db: "admin" },

{ role: "read", db: "local" }

Enabling Profiler

This is optional. Run the instance with the profiler or add profiling at the database level to monitor queries in QAN (not applicable for mongos).

To start at the instance level (enables profiling for all databases):

Shell

mongod <other options> --profile 2 --slowms 200 --rateLimit 100

or in mongod.conf:

mongod.conf profiling

Shell

operationProfiling:

mode: all

slowOpThresholdMs: 200

# (Below variable is available only with Percona Server for MongoDB.)

rateLimit: 100

To enable p[rofiling at DB level:

profiling db level

Shell

use dbname

db.setProfilingLevel(2)

Add MongoDB Instance to the pmm-client

Here use the same –cluster option name for all members from the same cluster and provide service-name to identify it:

add instance to pmm-client

Shell

sudo pmm-admin add mongodb \

--username=pmm_mongodb --password=password \

--query-source=profiler \

--cluster=mycluster \

--service-name=myc_mongoc2 \

--host=127.0.0.1 --port=37061

Check the Inventory Service

Then check whether the service was added successfully or not:

service check

Shell

$ sudo pmm-admin list

Service type Service name Address and port Service ID

MongoDB myc_mongoc2 127.0.0.1:37061 /service_id/02e261a1-e8e0-4eb4-8043-8616424500de

Agent type Status Metrics Mode Agent ID Service ID

pmm_agent Connected /agent_id/281b4046-4f4b-4897-bd2e-b771d3e97922

node_exporter Running push /agent_id/5e9b17a8-ecb9-47c3-8477-ce322047c4d9

mongodb_exporter Running push /agent_id/0067dd85-9a0a-47dd-976e-ae779deb982b /service_id/5c92f132-3005-45ab-84df-7541c286c34a

mongodb_profiler_agent Running /agent_id/18d3d87a-9bb9-48c1-8e3e-d8bae3f043bb /service_id/02e261a1-e8e0-4eb4-8043-8616424500de

From My Test

I used localhost to deploy the sharded cluster for the testing purpose as below:

Members list:

Shell

1 mongos (37050),

3 shards consist of 3 member replicaSet each (37051-37059),

3 config members(37060-37062)

Listing one mongod instance from the ps command:

Shell

balaguru@vinodh-UbuntuPC:~/mongodb/testshard$ ps -ef | grep mongod -w | head -1

balaguru 41883 2846 1 13:01 ? 00:04:04 mongod --replSet configRepl --dbpath /home/balaguru/mongodb/testshard/data/configRepl/rs1/db --logpath /home/balaguru/mongodb/testshard/data/configRepl/rs1/mongod.log --port 37060 --fork --configsvr --wiredTigerCacheSizeGB 1 --profile 2 --slowms 200 --rateLimit 100 --logappend

Adding mongodb services to pmm-admin:

pmm-admin add

Shell