Scribd
Upload
48 views16 pages

Replication

Replication allows for keeping identical copies of data on multiple servers. MongoDB uses replica sets, which consist of a primary server that accepts writes and secondary servers that copy the primary's data. If the primary fails, a secondary can be elected as the new primary. The members should have unique IPs and authentication/authorization enabled. Replica sets are limited to 50 members total, with only 7 allowed to vote for a new primary.

Uploaded by

srinu katams
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
48 views16 pages

Replication

Replication allows for keeping identical copies of data on multiple servers. MongoDB uses replica sets, which consist of a primary server that accepts writes and secondary servers that copy the primary's data. If the primary fails, a secondary can be elected as the new primary. The members should have unique IPs and authentication/authorization enabled. Replica sets are limited to 50 members total, with only 7 allowed to vote for a new primary.

Uploaded by

srinu katams
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Replication

What is replication
• Replication is a way of keeping identical copies of your data on
multiple servers
• With MongoDB, you set up replication by creating a replica set
• A replica set is a group of servers with one primary, the server
taking writes, and multiple secondaries, servers that keep copies of
the primary’s data
• If the primary crashes, the secondaries can elect a new primary
from amongst themselves.
• MongoDB Atlas provides replicasets for test, dev and production
Important points to consider

Secondary

Client Primary

Secondary
Important points to consider
• The members of the replica set should be started with ip other than
local ip.Use –bind_ip option
• Enable authentication and authorization
• rs is a global variable , containg methods to administrate replicat sets
• Replica sets are limited to 50 members in total and only 7 voting
members
• If you are creating a replica set that has more than seven
members, every additional member must be given zero votes
• > rs.add({"_id" : 7, "host" : “10.17.57.10:27025", "votes" : 0})
Setting up replication
• In this configuration , we have 3 servers. For simiplicity we are
starting all the three servers from the same VM, but on three
different ports
• Server lists
• 10.17.57.10 27018
• 10.17.57.10 27019
• 10.17.57.10 27020
• Create all the directories , required for data directory and log file
• $ mkdir /u01/data1 /u01/data2/ /u01/data3
Setting up replication
• Start all the three mongod process
• mongod --bind_ip localhost,10.17.57.10 --replSet rs0 --dbpath /u01/data1 --
port 27018 --oplogSize 200 --logpath /u01/data1/mongo.log --fork

• mongod --bind_ip localhost,10.17.57.10 --replSet rs0 --dbpath /u01/data2 --


port 27019 --oplogSize 200 --logpath /u01/data2/mongo.log --fork

• mongod --bind_ip localhost,10.17.57.10 --replSet rs0 --dbpath /u01/data3 --


port 27020 --oplogSize 200 --logpath /u01/data3/mongo.log --fork
Setting up replication
• Connect to one of the mongod process, and create a variable with member
information and replica set name . Make sure to use the replica set name
used while starting up the mongod servers
rconfig={
_id:"rs0",
members:[
{_id:0,host:"10.17.57.10:27018"},
{_id:1, host:"10.17.57.10:27019"},
{_id:2, host:"10.17.57.10:27020"}
]
}
• Initiate the replication configuration
• > rs.initiate(rconfig)
Setting up replication
• Replica set will elect a primary soon
• The mongo shell prompt displays if its primary or secondary
• > rs.status() will show the status of the replication configuration
• rs0:PRIMARY> indicates , you are connected to the primary server of “rs0”,
replication configuration
• > db.isMaster() helps to determine if we have connection to the primary .
Mostly used in the programing
• Both commands show the list of all the members of the replica set
Reading from the secondary
• By default reading from secondary is not enabled.
• "errmsg" : "not master and slaveOk=false",
• To solve
• Connect to secondary
• > rs.secondaryOk()
• Try to select documents which were inserted on primary
• > use movies
• > db.movies.find()
Testing the failover secondary
• Connect to primary and shut it down
• use admin
• > db.shutdownServer()
• Connect to any other secondary
• > rs.status()
• Look for “"stateStr" : "PRIMARY“”, to understand which is the current primary
Adding & Removing members
• Start a new mongod
• mongod --bind_ip localhost,10.17.57.10 --replSet rs0 --dbpath /u01/data4 --
port 27021 --oplogSize 200 --logpath /u01/data4/mongo.log –fork
• Connect to primary
• Add the new member
• > rs.add("10.17.57.10:27021")
• Check the status of the configuration
• > rs.config()
• Remove a member
• > rs.remove("10.17.57.10:27021")
Reconfiguring Replicaset
• Lets assume, that one of the member is starting on a new port. To solve thi , we have
to change the port number of that member.
• Stop the member , and restart using the new port number
• lastHeartbeatMessage" : "Error connecting to 10.17.57.10:27021 :: caused by :: Connection
refused", // rs.status displays this message
• Modify the member with new port number
• > var conf=rs.config()
• > conf.members[3].host="10.17.57.10:27022“
• Reconfigure the replica set
• > rs.reconfig(conf)
• Check the configuration and status
• > rs.config()
• > rs.status()
•
Considerations while designing replicaset
• Design configuration such a way that primary can reach more than half of the
members
• Majority is designed by the number of members in the configuration , not based on
the available members.
• Members will ping other members every 2 second.
• If a heartbeat does not return from a member within 10 seconds, the other
members mark the delinquent member as inaccessible.
• If the member unable to reach the primary in specified time, it will propose itself as
primary
• Other members after receiving the new primary request checks the following.
• Are they able to reach old primary
• Is the member who request to primary has the latest transaction
• Priority of the member
States that indicate a problem in replication
• DOWN
• If a member was up but then becomes unreachable
• UNKNOWN
• If a member has never been able to reach another member, it will not know what state
it’s in, so it will report it as UNKNOWN.
• REMOVED
• This is the state of a member that has been removed from the set.
• ROLLBACK
• This state is used when a member is rolling back data,
Different States in Replication
• STARTUP
• This is the state a member is in when it’s first started, while MongoDB is attempting to
load its replica set configuration. Once the configuration has been loaded, it transitions
to STARTUP2.
• STARTUP2
• This state lasts throughout the initial sync process, which typically takes just a few
seconds. The member forks off a couple of threads to handle replication and elections
and then transitions into the next state: RECOVERING.
• RECOVERING
• This state indicates that the member is operating correctly but is not available for reads
• ARBITER
• Arbiters have a special state and should always be in this state during normal
operation.
Increase the size of Ops log
• Connect to the replica set member whose Ops log size to be increased.
• Verify the current size of the oplog:
• > use local
• > db.oplog.rs.stats(1024*1024).maxSize
• Change the oplog size of the replica set member:
• > db.adminCommand({replSetResizeOplog: 1, size: 16000})

You might also like