Running MongoDB on OpenEBS

This section provides detailed instructions which allow you to perform the following tasks.

  • Run a MongoDB StatefulSet on OpenEBS storage in a Kubernetes cluster
  • Generate standard OLTP load on MongoDB using a custom Sysbench tool
  • Test data replication across the MongoDB instances

1. Run OpenEBS Operator

For the prerequisites and running OpenEBS Operator, see Running OpenEBS Operator.

2. Deploy Mongo-StatefulSet with OpenEBS Storage

Use OpenEBS as persistent storage for the MongoDB StatefulSet by selecting an OpenEBS storage class in the persistent volume claim. A sample MongoDB SatefulSet yaml (with container attributes and pvc details) is available in the OpenEBS git repository.

The number of replicas in the StatefulSet can be modified as required. The following example uses three replicas. The replica count can be edited in the StatefulSet specification.

apiVersion: apps/v1beta1
kind: StatefulSet
 name: mongo
 serviceName: "mongo"
 replicas: 3
       role: mongo
       environment: test

Apply the mongo-S yaml using the following commands.

test@Master:~$ kubectl apply -f mongo-S.yml
service "mongo" created
S "mongo" created

Verify that MongoDB replicas, mongo headless service and OpenEBS persistent volumes comprising of the controller and replica pods are successfully deployed and are in Running state.

test@Master:~$ kubectl get pods
NAME                                                             READY     STATUS    RESTARTS   AGE
maya-apiserver-1089964587-x5q15                                  1/1       Running   0          8m
mongo-0                                                          2/2       Running   0          2m
mongo-1                                                          2/2       Running   0          2m
mongo-2                                                          2/2       Running   0          1m
openebs-provisioner-1149663462-5pdcq                             1/1       Running   0          8m
pvc-0d39583c-bad7-11e7-869d-000c298ff5fc-ctrl-4109100951-v2ndc   1/1       Running   0          2m
pvc-0d39583c-bad7-11e7-869d-000c298ff5fc-rep-1655873671-50f8z    1/1       Running   0          2m
pvc-0d39583c-bad7-11e7-869d-000c298ff5fc-rep-1655873671-ctp0q    1/1       Running   0          2m
pvc-21da76b6-bad7-11e7-869d-000c298ff5fc-ctrl-2618026111-z5hzt   1/1       Running   0          2m
pvc-21da76b6-bad7-11e7-869d-000c298ff5fc-rep-187343257-9w46n     1/1       Running   0          2m
pvc-21da76b6-bad7-11e7-869d-000c298ff5fc-rep-187343257-sd5hl     1/1       Running   0          2m
pvc-3a9ca1ec-bad7-11e7-869d-000c298ff5fc-ctrl-2347166037-vsc2t   1/1       Running   0          1m
pvc-3a9ca1ec-bad7-11e7-869d-000c298ff5fc-rep-849715916-3w1c7     1/1       Running   0          1m
pvc-3a9ca1ec-bad7-11e7-869d-000c298ff5fc-rep-849715916-f2f3p     1/1       Running   0          1m

test@Master:~$ kubectl get svc
NAME                                                CLUSTER-IP       EXTERNAL-IP PORT(S)             AGE
kubernetes                                        <none>      443/TCP             19h
maya-apiserver-service                       <none>      5656/TCP            8m
mongo                                               None             <none>      27017/TCP           3m
pvc-0d39583c-bad7-11e7-869d-000c298ff5fc-ctrl-svc     <none>      3260/TCP,9501/TCP   3m
pvc-21da76b6-bad7-11e7-869d-000c298ff5fc-ctrl-svc   <none>      3260/TCP,9501/TCP   2m
pvc-3a9ca1ec-bad7-11e7-869d-000c298ff5fc-ctrl-svc    <none>      3260/TCP,9501/TCP   1m


It may take some time for the pods to start as the images must be pulled and instantiated. This is also dependent on the network speed.

3. Generate Load on the MongoDB Instance

In this example, you will be using a custom-built Sysbench framework integrated with support for OLTP tests MongoDB through the lua scripts. Sysbench is a multi-purpose benchmarking tool which can run DB benchmarks as well as regular raw/file device input/output.

Installing Sysbench

The following procedure helps you install Sysbench.

  • Download the appropriate branch of Percona-Lab’s Sysbench fork with support for MongoDB integration on Kubernetes nodes. The Sysbench dependencies are installed in these Kubernetes nodes. (see, `Prerequisites`_)
git clone -b dev-mongodb-support-1.0
  • Enter the Sysbench local repository and perform the following commands in the given order.
cd sysbench



In case of errors where some header files belonging to the libbson/libmongoc packages are not found, update the include path. A workaround for this is to place all header files inside libbson-1.0 and libmongoc-1.0 into /usr/include folder.

Executing the Sysbench Benchmark

  • Identify the primary MongoDB instance name or IP (In the current S specification YAML, “mongo-0” is always configured as the primary instance that takes the client input/output)
  • Trigger the Sysbench command using the following command to -
    • prepare the database
    • add the collections
    • perform the benchmark run

Note: Replace the mongo-url parameter based on the appropriate IP which can be obtained by kubectl describe pod mongo-0 | grep IP

test@Host02:~/sysbench$ ./sysbench/sysbench --mongo-write-concern=1 --mongo-url="mongodb://" --mongo-database-name=sbtest --test=./sysbench/tests/mongodb/oltp.lua --oltp_table_size=100 --oltp_tables_count=10 --num-threads=10 --rand-type=pareto --report-interval=10 --max-requests=0 --max-time=600 --oltp-point-selects=10 --oltp-simple-ranges=1 --oltp-sum-ranges=1 --oltp-order-ranges=1 --oltp-distinct-ranges=1 --oltp-index-updates=1 --oltp-non-index-updates=1 --oltp-inserts=1 run

The parameters used for Sysbench can be modified based on system capability and storage definition to obtain realistic benchmark figures.

The benchmark output displayed is similar to the following:

sysbench 1.0:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 10
Report intermediate results every 10 second(s)
Initializing random number generator from current time

Initializing worker threads...

setting write concern to 1
Threads started!

[  10s] threads: 10, tps: 56.60, reads: 171.50, writes: 170.40, response time: 316.14ms (95%), errors: 0.00, reconnects:  0.00
[  20s] threads: 10, tps: 74.70, reads: 222.90, writes: 223.50, response time: 196.30ms (95%), errors: 0.00, reconnects:  0.00
[  30s] threads: 10, tps: 76.00, reads: 227.70, writes: 228.00, response time: 196.71ms (95%), errors: 0.00, reconnects:  0.00
[  40s] threads: 10, tps: 79.60, reads: 239.70, writes: 238.80, response time: 329.08ms (95%), errors: 0.00, reconnects:  0.00
OLTP test statistics:
  queries performed:
      read:                            154189
      write:                           154122
      other:                           51374
      total:                           359685
  transactions:                        51374  (85.61 per sec.)
  read/write requests:                 308311 (513.79 per sec.)
  other operations:                    51374  (85.61 per sec.)
  ignored errors:                      0      (0.00 per sec.)
  reconnects:                          0      (0.00 per sec.)

General statistics:
    total time:                          600.0703s
    total number of events:              51374
    total time taken by event execution: 6000.1853s
    response time:
         min:                                 26.11ms
         avg:                                116.79ms
         max:                               2388.03ms
         approx.  95 percentile:             224.00ms

Threads fairness:
    events (avg/stddev):           5137.4000/21.50
    execution time (avg/stddev):   600.0185/0.02

While the benchmark is in progress, performance and capacity usage statistics on the OpenEBS storage volume can be viewed using the mayactl commands that must be executed on the maya-apiserver pod.

Run an interactive bash session for the maya-apiserver pod container.

test@Master:~$ kubectl exec -it maya-apiserver-1089964587-x5q15 /bin/bash

Obtain the list of OpenEBS persistent volumes created by the MongoDB S application YAML.

root@maya-apiserver-1089964587-x5q15:/# maya volume list
Name                                      Status
pvc-0d39583c-bad7-11e7-869d-000c298ff5fc  Running
pvc-21da76b6-bad7-11e7-869d-000c298ff5fc  Running

View usage and input/output metrics for the required volume through the stats command.

root@maya-apiserver-1089964587-x5q15:/# maya volume stats pvc-0d39583c-bad7-11e7-869d-000c298ff5fc
IQN     :
Volume  : pvc-0d39583c-bad7-11e7-869d-000c298ff5fc
Portal  :
Size    : 5G

     Replica|   Status|   DataUpdateIndex|
            |         |                  ||   Online|              4341||   Online|              4340|

----------- Performance Stats -----------

   r/s|   w/s|   r(MB/s)|   w(MB/s)|   rLat(ms)|   wLat(ms)|
     0|    14|     0.000|    14.000|      0.000|     71.325|

------------ Capacity Stats -------------

   Logical(GB)|   Used(GB)|
         0.214|      0.205|

Verifying MongoDB Replication

  • Log in to the primary instance of the MongoDB S through the in-built Mongo shell and verify that the sbtest test database is created by Sysbench in the previous procedure.
test@Master:~$ kubectl exec -it mongo-0 /bin/bash
root@mongo-0:/# mongo

MongoDB shell version v3.4.9
connecting to: mongodb://
MongoDB server version: 3.4.9
rs0:PRIMARY> show dbs
admin   0.000GB
local   0.006GB
sbtest  0.001GB
  • Run the replication status command on the master/primary instance of the S. In the output, verify that the values (timestamps) for optimeDate on both members are similar.
   rs0:PRIMARY> rs.status()
  "set" : "rs0",
  "date" : ISODate("2017-10-23T07:26:36.679Z"),
  "myState" : 1,
  "term" : NumberLong(1),
  "heartbeatIntervalMillis" : NumberLong(2000),
  "optimes" : {
          "lastCommittedOpTime" : {
                  "ts" : Timestamp(1508743595, 51),
                  "t" : NumberLong(1)
          "appliedOpTime" : {
                  "ts" : Timestamp(1508743596, 40),
                  "t" : NumberLong(1)
          "durableOpTime" : {
                  "ts" : Timestamp(1508743595, 71),
                  "t" : NumberLong(1)
  "members" : [
                  "_id" : 0,
                  "name" : "",
                  "health" : 1,
                  "state" : 1,
                  "stateStr" : "PRIMARY",
                  "uptime" : 243903,
                  "optime" : {
                          "ts" : Timestamp(1508743596, 40),
                          "t" : NumberLong(1)
                  "optimeDate" : ISODate("2017-10-23T07:26:36Z"),
                  "electionTime" : Timestamp(1508499738, 2),
                  "electionDate" : ISODate("2017-10-20T11:42:18Z"),
                  "configVersion" : 5,
                  "self" : true
                  "_id" : 1,
                  "name" : "",
                  "health" : 1,
                  "state" : 2,
                  "stateStr" : "SECONDARY",
                  "uptime" : 243756,
                  "optime" : {
                          "ts" : Timestamp(1508743595, 51),
                          "t" : NumberLong(1)
                  "optimeDurable" : {
                          "ts" : Timestamp(1508743595, 34),
                          "t" : NumberLong(1)
                  "optimeDate" : ISODate("2017-10-23T07:26:35Z"),
                  "optimeDurableDate" : ISODate("2017-10-23T07:26:35Z"),
                  "lastHeartbeat" : ISODate("2017-10-23T07:26:35.534Z"),
                  "lastHeartbeatRecv" : ISODate("2017-10-23T07:26:34.894Z"),
                  "pingMs" : NumberLong(6),
                  "syncingTo" : "",
                  "configVersion" : 5
                  "_id" : 2,
                  "name" : "",
                  "health" : 1,
                  "state" : 2,
                  "stateStr" : "SECONDARY",
                  "uptime" : 243700,
                  "optime" : {
                          "ts" : Timestamp(1508743595, 104),
                          "t" : NumberLong(1)
                  "optimeDurable" : {
                          "ts" : Timestamp(1508743595, 34),
                          "t" : NumberLong(1)
                  "optimeDate" : ISODate("2017-10-23T07:26:35Z"),
                  "optimeDurableDate" : ISODate("2017-10-23T07:26:35Z"),
                  "lastHeartbeat" : ISODate("2017-10-23T07:26:35.949Z"),
                  "lastHeartbeatRecv" : ISODate("2017-10-23T07:26:35.949Z"),
                  "pingMs" : NumberLong(0),
                  "syncingTo" : "",
                  "configVersion" : 5
  "ok" : 1
  • You could further confirm the presence of the database with the same size on secondary instances (for example, mongo-1).


By default, the databases cannot be viewed on the secondary instance through the show dbs command, unless we set the slave context.

rs0:SECONDARY> rs.slaveOk()

rs0:SECONDARY> show dbs
admin   0.000GB
local   0.005GB
sbtest  0.001GB
  • The time lag between the MongoDB instances can be found using the following command, which can be executed on either instance.
rs0:SECONDARY> rs.printSlaveReplicationInfo()
     syncedTo: Mon Oct 23 2017 07:28:27 GMT+0000 (UTC)
     0 secs (0 hrs) behind the primary
     syncedTo: Mon Oct 23 2017 07:28:27 GMT+0000 (UTC)
     0 secs (0 hrs) behind the primary