Replication

When distributing data across several instances, Memgraph uses replication to provide a satisfying ratio of the following properties, known from the CAP theorem:

1. Consistency (C) - every node has the same view of data at a given point in time
2. Availability (A) - all clients can find a replica of the data, even in the case of a partial node failure
3. Partition tolerance (P) - the system continues to work as expected despite a partial network failure

In the replication process, the data is replicated from one storage (MAIN instance) to another (REPLICA instances).

Data replication implementation basics

In Memgraph, all instances are MAIN upon starting. When creating a replication cluster, one instance has to be chosen as the MAIN instance. The rest of the instances have to be demoted to REPLICA roles and have a port defined using a Cypher query.

If you want instances to remember their role and configuration in a replication cluster upon restart, they need to be initialized with the --replication-restore-state-on-startup set to true and remain true throughout the instances' lifetime. Otherwise and by default, restarted instances will start as MAIN instances disconnected from any replication cluster.

Once demoted to REPLICA instances, they will no longer accept write queries. In order to start the replication, each REPLICA instance needs to be registered from the MAIN instance by setting a replication mode (SYNC or ASYNC) and specifying the REPLICA instance's socket address.

The replication mode defines the terms by which the MAIN instance can commit the changes to the database, thus modifying the system to prioritize either consistency or availability:

• SYNC - After committing a transaction, the MAIN instance will communicate the changes to all REPLICA instances running in SYNC mode and wait until it receives a response or information that a timeout is reached. SYNC mode ensures consistency and partition tolerance (CP), but not availability for writes. If the primary database has multiple replicas, the system is highly available for reads. But, when a replica fails, the MAIN instance can't process the write due to the nature of synchronous replication.

• ASYNC - The MAIN instance will commit a transaction without receiving confirmation from REPLICA instances that they have received the same transaction. ASYNC mode ensures system availability and partition tolerance (AP), while data can only be eventually consistent.

Once the REPLICA instances are registered, data storage of the MAIN instance is replicated and synchronized using transaction timestamps and durability files (snapshot files and WALs). Memgraph does not support replication of authentication configurations, query and authentication modules, and audit logs.

By using the timestamp, the MAIN instance knows the current state of the REPLICA. If the REPLICA is not synchronized with the MAIN instance, the MAIN instance sends the correct data for synchronization kept as deltas within WAL files. Deltas are the smallest possible updates of the database, but they carry enough information to synchronize the data on a REPLICA. Memgraph stores only remove actions as deltas, for example, REMOVE key:value ON node_id.

If the REPLICA is so far behind the MAIN instance that the synchronization using WAL files and deltas within it is impossible, Memgraph will use snapshots to synchronize the REPLICA to the state of the MAIN instance.

From Memgraph version 2.15, a REPLICA instance has integrated support to only listen to one MAIN. This part is introduced to support the high availability but also reflects on the replication. The mechanism that is used is a unique identifier that which MAIN instance sends to all REPLICAs when REPLICA is first registered on a MAIN. A REPLICA stores the UUID of the MAIN instance it listens to. The MAIN's UUID is also stored on a disk, in case of restart of an instance to continue listening to the correct MAIN instance. When REPLICA restarts, --replication-restore-state-on-startup must be set to true to continue getting updates from the MAIN.

Auth data replication (Enterprise)

If you are using a Memgraph Enterprise license, all authentication/authorization data, including users, roles, and associated permissions, will be replicated.

Auth modules replication (Enterprise)

Authentication modules are not replicated and must be configured manually by the administrator.

Multi-tenant data replication (Enterprise)

When you are using a Memgraph Enterprise license, multi-tenant commands are replicated as any other data command. Database manipulation is allowed only on MAIN. However, REPLICAs have the ability to use databases and read data contained in them.

When dropping a database used on a REPLICA, the REPLICA will receive the command and will partially drop the database. It will hide the database and prevent any new usage. Once all clients have released the database, it will be deleted entirely.

Running multiple instances

When running multiple instances, each on its own machine, run Memgraph as you usually would.

If you are exploring replication and running multiple instances on one machine, you can run Memgraph with Docker, but if you are using volumes, they need to be called differently and each instance needs to be exposed via a different port.

Check the example of creating a replication cluster.

Assigning roles

Each Memgraph instance has the role of the MAIN instance when it is first started.

Also, by default, each crashed instance restarts with its previous role (MAIN as MAIN, REPLICA as REPLICA). To change this behavior, set the --replication-restore-state-on-startup to false when first initializing the instance. In this way, all instances will get restarted as MAIN.

Assigning the REPLICA role

Once you decide what instance will be the MAIN instance, all the other instances that will serve as REPLICA instances need to be demoted and have the port set using the following query:

SET REPLICATION ROLE TO REPLICA WITH PORT <port_number>;

If you set the port of each REPLICA instance to 10000, it will be easier to register replicas later on because the query for registering replicas uses a port 10000 as the default one.

Otherwise, you can use any unassigned port between 1000 and 10000.

Assigning the MAIN role

The replication cluster should only have one MAIN instance in order to avoid errors in the replication system. If the original MAIN instance fails, you can promote a REPLICA instance to be the new MAIN instance by running the following query:

SET REPLICATION ROLE TO MAIN;

If the original instance was still alive when you promoted a new MAIN, you need to resolve any conflicts and manage replication manually.

If you demote the new MAIN instance back to the REPLICA role, it will not retrieve its original function. You need to drop it from the MAIN and register it again.

If the crashed MAIN instance goes back online once a new MAIN is already assigned, it cannot reclaim its previous role. It can be cleaned and demoted to become a REPLICA instance of the new MAIN instance.

Checking the assigned role

To check the replication role of an instance, run the following query:

SHOW REPLICATION ROLE;

Registering REPLICA instances

Once all the nodes in the cluster are assigned with appropriate roles, you can enable replication in the MAIN instance by registering REPLICA instances, setting a replication mode (SYNC and ASYNC), and specifying the REPLICA instance's socket address. Memgraph doesn't support chaining REPLICA instances, that is, a REPLICA instance cannot be replicated on another REPLICA instance.

If you want to register a REPLICA instance with a SYNC replication mode, run the following query:

REGISTER REPLICA name SYNC TO <socket_address>;

If you want to register a REPLICA instance with an ASYNC replication mode, run the following query:

REGISTER REPLICA name ASYNC TO <socket_address>;

The socket address must be a string value as follows:

"IP_ADDRESS:PORT_NUMBER"

where IP_ADDRESS is a valid IP address, and PORT_NUMBER is a valid port number, for example:

"172.17.0.4:10050"

The default value of the PORT_NUMBER is 10000, so if you set REPLICA roles using that port, you can define the socket address specifying only the valid IP address:

"IP_ADDRESS"

Example of a <socket_address> using only IP_ADDRESS:

"172.17.0.5"

Also, you can register REPLICA instances using DNS names. In that case, the socket address must be a string value as follows:

"DOMAIN_NAME:PORT_NUMBER"

where DOMAIN_NAME is a valid domain name, and PORT_NUMBER is a valid port number, for example:

"memgraph-replica.memgraph.net:10050"

If you set REPLICA roles using port 10000, you can define the socket address specifying only the valid domain name, for example:

"memgraph-replica.memgraph.net"

When a REPLICA instance is registered, it will start replication in ASYNC mode until it synchronizes to the current state of the database. Upon synchronization, REPLICA instances will either continue working in the ASYNC mode or reset to SYNC mode.

Listing all registered REPLICA instances

You can check all the registered REPLICA instances and their details by running the following query:

SHOW REPLICAS;

Dropping a REPLICA instance

To drop a replica, run the following query:

DROP REPLICA <name>;

MAIN and REPLICA synchronization

By comparing timestamps, the MAIN instance knows when a REPLICA instance is not synchronized and is missing some earlier transactions. The REPLICA instance is then set into a RECOVERY state, where it remains until it is fully synchronized with the MAIN instance#synchronizing-instances.

The missing data changes can be sent as snapshots or WAL files. Snapshot files represent an image of the current state of the database and are much larger than the WAL files, which only contain the changes, deltas. Because of the difference in file size, Memgraph favors the WAL files. It is important to note that replicas receive only changes which are made durable on the MAIN instance, in other words changes which are alredy fsynced.

While the REPLICA instance is in the RECOVERY state, the MAIN instance calculates the optimal synchronization path based on the REPLICA instance's timestamp and the current state of the durability files while keeping the overall size of the files necessary for synchronization to a minimum.

Set up a replication cluster

In the replication process, the data is replicated from one storage (MAIN instance) to another (REPLICA instances), thus providing a combination of consistency, availability and partition tolerance when distributing data over several instances.

This example demonstrates how to create a simple cluster of nodes running Memgraph instances, and set up replication using various replication modes.

Cluster topology

The cluster will consist of three nodes, one MAIN instance and two REPLICA instances. In order to showcase the creation of REPLICA instances with different replication modes, we will create:

• The MAIN instance - contains the original data that will be replicated to REPLICA instances
• REPLICA instance 1 - replication in the SYNC mode
• REPLICA instance 2 - replication in the ASYNC mode

Run multiple instances

If you are running multiple instances, each on its own machine, run Memgraph as you usually would.

If you are exploring replication and running multiple instances on one machine, you need expose different ports for each instance.

The MAIN instance:

docker run -p 7687:7687 memgraph/memgraph-mage --data-recovery-on-startup=true

REPLICA instance 1:

docker run -p 7688:7687 memgraph/memgraph-mage --data-recovery-on-startup=true

REPLICA instance 2:

docker run -p 7689:7687 memgraph/memgraph-mage --data-recovery-on-startup=true

You can connect to each instance using the Memgraph Lab desktop application, or any other external application by changing the port:

• the MAIN instance - localhost:7687
• REPLICA instance 1 - localhost:7688
• REPLICA instance 2 - localhost:7689

If you need to define volumes, each volume needs to be called differently.

Demote an instance to a REPLICA role

Run the following query in both REPLICA instances to demote them to the REPLICA role:

SET REPLICATION ROLE TO REPLICA WITH PORT 10000;

If you set the port of each REPLICA instance to 10000, it will be easier to register replicas later on because the query for registering replicas uses port 10000 as the default one.

Otherwise, you can use any unassigned port between 1000 and 10000.

Register REPLICA instances

To register a REPLICA instance, you need to find out the IP address of each instance.

The IP addresses will probably be:

• the MAIN instance - 172.17.0.2
• REPLICA instance 1 - 172.17.0.3
• REPLICA instance 2 - 172.17.0.4

If they are not, please change the IP addresses in the following queries to match the IP addresses on your cluster.

Then, run the following queries from the MAIN instance to register REPLICA instances:

1. REPLICA instance 1 at 172.17.0.3

   REGISTER REPLICA REP1 SYNC TO "172.17.0.3";

   REPLICA instance 1 is called REP1, its replication mode is SYNC, and it is located at IP address 172.17.0.3. with port 10000.

   Once the MAIN instance commits a transaction, it will communicate the changes to all REPLICA instances running in SYNC mode and wait until it receives a response that the changes have been applied to the REPLICAs or that a timeout has been reached.

   If you used any port other than 10000 while demoting a REPLICA instance, you will need to specify it like this: "172.17.0.3:5000"

2. REPLICA instance 2 at 172.17.0.4

   REGISTER REPLICA REP2 ASYNC TO "172.17.0.4";

   REPLICA instance 2 is called REP2, its replication mode is ASYNC, and it is located at IP address 172.17.0.4. with port 10000.

   When the REPLICA instance is running in ASYNC mode, the MAIN instance will commit a transaction without receiving confirmation from REPLICA instances that they have received the same transaction. ASYNC mode ensures system availability and partition tolerance.

   If you used any port other than 10000 while demoting a REPLICA instance, you will need to specify it like this: "172.17.0.4:5000"

Check info about registered REPLICA instances

Check REPLICA instances by running the following query from the MAIN instance:

SHOW REPLICAS;

The result has information regarding each individual replica:

1. replica's name
2. address
3. type (sync/async)
4. system information
5. multi-tenant information (for each database, we provide the current timestamp, how many tick is the replica's version behind and the current status)

Underlying implementation

Uninterrupted data and operational availability in production systems are critical and can be achieved in many ways. In Memgraph we opted for replication.

In distributed systems theory the CAP theorem, also named Brewer's theorem, states that any distributed system can simultaneously guarantee two out of the three properties:

1. Consistency (C) - every node has the same view of data at a given point in time
2. Availability (A) - all clients can find a replica of the data, even in the case of a partial node failure
3. Partition tolerance (P) - the system continues to work as expected despite a partial network failure

Most of the Memgraph use cases do not benefit from well-known algorithms that strive to achieve all three CAP properties, such as Raft, because due to their complexity, they produce performance issues. Memgraph use-cases are based on running analytical graph workloads on real-time data, demanding a simpler concept such as replication.

Replication consists of replicating data from one storage to one or several other storages. The downside of its simplicity is that only two out of three CAP properties can be achieved.

Replication implementation in Memgraph

To enable replication, there must be at least two instances of Memgraph in a cluster. Each instance has one of two roles: MAIN or REPLICA. The MAIN instance accepts read and write queries to the database and REPLICA instances accept only read queries.

The changes or state of the MAIN instance are replicated to the REPLICA instances in a SYNC or ASYNC mode. The SYNC mode ensures consistency and partition tolerance (CP), but not availability for writes. The ASYNC mode ensures system availability and partition tolerance (AP), while data can only be eventually consistent.

By using the timestamp, the MAIN instance knows the current state of the REPLICA. If the REPLICA is not synchronized with the MAIN instance, the MAIN instance sends the correct data for synchronization as WAL files.

If the REPLICA is so far behind the MAIN instance that the synchronization using WAL files is impossible, Memgraph will use snapshots.

Replication modes

Replication mode defines the terms by which the MAIN instance can commit the changes to the database, thus modifying the system to prioritize either consistency or availability. There are two possible replication modes implemented in Memgraph replication:

• SYNC
• ASYNC

When a REPLICA instance is registered and added to the cluster, it will start replicating in ASYNC mode. That will allow it to catch up to the current state of the MAIN instance. When the REPLICA instance synchronizes with the MAIN instance, the replication mode will change according to the mode defined during registration.

SYNC replication mode

SYNC mode is the most straightforward replication mode in which the main storage thread waits for the response and cannot continue until the response is received or a timeout is reached.

The following diagrams express the behavior of the MAIN instance in cases when SYNC REPLICA doesn't answer within the expected timeout.

SYNC REPLICA going down when creating index, uniqueness constraint or existence constraint

SYNC REPLICA going down when dropping index, uniqueness constraint or existence constraint

SYNC REPLICA going down adding/updating/deleting data

ASYN replication mode

In the ASYNC replication mode, the MAIN instance will commit a transaction without receiving confirmation from REPLICA instances that they have received the same transaction. This means that the MAIN instance does not wait for the response from the REPLICA instances in the main thread but in some other thread.

A new thread can be created every time a transaction needs to be replicated to the REPLICA instance, but because transactions are committed often and use a lot of resources, each REPLICA instance has one permanent thread connecting it with the MAIN instance. Using this background thread, the MAIN instance pushes replication tasks to the REPLICA instance, creates a custom thread pool pattern, and receives confirmations of successful replication from the REPLICATION instance.

ASYNC mode ensures system availability and partition tolerance.

Synchronizing instances

By comparing timestamps, the MAIN instance knows when a REPLICA instance is not synchronized and is missing some earlier transactions. The REPLICA instance is then set into a RECOVERY state, where it remains until it is fully synchronized with the MAIN instance.

The missing data changes can be sent as snapshots or WAL files. Snapshot files represent an image of the current state of the database and are much larger than the WAL files, which only contain the changes, deltas. Because of the difference in file size, Memgraph favors the WAL files.

While the REPLICA instance is in the RECOVERY state, the MAIN instance calculates the optimal synchronization path based on the REPLICA instance's timestamp and the current state of the durability files while keeping the overall size of the files necessary for synchronization to a minimum.

Imagine there were 5 changes made to the database. Each change is saved in a WAL file, so there are 5 WAL files, and the snapshot was created after 2 changes. The REPLICA instance can be synchronized using a snapshot and the 3 latest WAL files or using 5 WAL files. Both options would correctly synchronize the instances, but 5 WAL files are much smaller.

The durability files are constantly being created, deleted, and updated. Also, each replica could need a different set of files to sync. There are several ways to ensure that the necessary files persist and that instances can read the WAL files currently being updated without affecting the performance of the rest of the database.

Locking durability files

Durability files are also used for recovery and are periodically deleted to eliminate redundant data. The problem is that they can be deleted while they are being used to synchronize a REPLICA with the MAIN instance.

To delay the file deletion, Memgraph uses a file retainer that consists of multiple lockers. Threads can store and lock the files they found while searching for the optimal recovery path in the lockers, thus ensuring the files will still exist once they are sent to the REPLICA instance as a part of the synchronization process. If another part of the system sends a deletion request for a certain file, the file retainer first checks if that file is locked in a locker. If it is not, it is deleted immediately. If the file is locked, the file retainer adds the file to the deletion queue. The file retainer will periodically clean the queue by deleting the files that are no longer locked inside the locker.

Writing and reading files simultaneously

Memgraph internal file buffer is used when writing deltas to WAL files, and mid-writing, the content of one WAL file can be divided across two locations. If at that point that WAL file is used to synchronize the REPLICA instance, once the data is being read from the internal buffer, the buffer can be flushed, and the REPLICA could receive an invalid WAL file because it is missing a chunk of data. It could also happen that the WAL file is sent before all the transactions are written to the internal buffer.

To avoid these issues, flushing of that internal buffer is disabled while the current WAL is sent to a REPLICA instance. To get all the data necessary for the synchronization, the replication thread reads the content directly from the WAL file, then reads how many bytes are written in the buffer and copies the data to another location. Then the flushing is enabled again, and the transaction is replicated using the copied buffer. Because the access to the internal buffer was not blocked, new data can be written. The content of the buffer (including any new data) is then written in a new WAL file that will be sent in the next synchronization process.

Fixing timestamp consistency

Timestamps are used to compare the state of the REPLICA instance in comparison to the MAIN instance.

At first, we used the current timestamp without increasing its value for global operations, like creating an index or creating a constraint. By using a single timestamp, it was impossible to know which operations the REPLICA had applied because sequential global operations had the same timestamp. To avoid this issue, a unique timestamp is assigned to each global operation.

As replicas allow read queries, each of those queries was assigned with its own timestamp. Those timestamps caused issues when the replicated write transactions were assigned an older timestamp. A read transaction would return different data from the same read query if a transaction was replicated between those two read transactions which obstructed the snapshot isolation. To avoid this problem, the timestamp on REPLICA instances isn't increased because the read transactions don't produce any changes, so no deltas need to be timestamped.

Incompatible instances

To avoid issues when the durability files of two different database instances are stored in the same folder, a unique ID is assigned to each storage instance. The same ID is then assigned to the durability files. Replication uses the instance ID to validate that the files and the database are compatible.

A unique ID epoch_id is also assigned each time an instance is run as the MAIN instance in the replication cluster to check if the data is compatible for replication. The epoch_id is necessary when the original MAIN instance fails, a REPLICA instance becomes a new MAIN, and after some time, the original MAIN instance is brought back online. If no transactions were run on the original MAIN instance, the difference in timestamps will indicate that it is behind the new MAIN, and it would be impossible to set the original MAIN-REPLICA relationship. But if the transactions were run on the original MAIN after it was brought back online, the timestamp would be of no help, but the epoch_id would indicate incomparability, thus preventing the original MAIN from reclaiming its original role.