Initialize Inter-Sync Gateway Replications

    +

    Initializing and running inter-Sync Gateway replication

    Related topics: Overview | Run | Manage | Monitor | Conflict

    Context Clarification

    This content relates only to inter-Sync Gateway replication in Sync Gateway 2.8+. For documentation on pre-2.8 inter-Sync Gateway replication (also known as SG Replicate) — see the documentation for the appropriate release.

    Introduction

    Replications are initialized by submitting a replication definitionglossary icon using either:

    • A 'JSON' configuration file (sync-gateway-config.json)

    • The Admin REST API, using a utility such as curl, or an application such as Postman.

    Wherever they are defined, the elements of a replication definition are the same, with the exception of the adhoc Admin REST API endpoint used to specify that the replication is ad hoc [1].

    Example 1. Replication Characteristics Highlights
    • Replication highlights

    • Running highlights

    • There are two types of replication: persistent and ad hoc (REST API only).

    • Replications of both types can run in one-shot or continuous replications modes.

    • All replications involve at least one local database.

    • Replications can be configured to purge documents when channel access is revoked (a removal notification is received).

    • Persistent continuous replications can be:

      • Reset — a checkpointglossary icon can be reset to zero

      • Updated — only the parameter values provided in the PUT request body will be updated

    • Persistent and ad hoc replications can be:

      • Removed — only the replication_id is needed to delete ongoing continuous or one-shot replications.

    • ENTERPRISE EDITION only:

      • Replications can use delta-sync mode, whereby only the changed data-items are replicated.

    • Multiple identical replicators can be initiated on a Sync Gateway node provided each has a unique replication_Id.

    • inter-Sync Gateway replications introduced in Sync Gateway 2.8 as well as SG-Replicate can run on the same node, but you must ensure that they each have a different replication_id.

    • The user under which replication is being run must have read and write access to the data being replicated.

    • Exponential backoff when connection lost; this can be customized using the max_backoff_time configuration setting.

    • replications will continue trying to connect for 30 minutes following authentication failure (including user-invalid/doesn’t exist).

    • Running replications can be stopped. Stopped replications can be (re)Started.

    • If ALL the Sync Gateway nodes in a source or target Sync Gateway cluster go down in the middle of continuous replication, by default, the system should pick up from the last document that was successfully processed by both sides when the replication/cluster is restarted

    • REST ONLY

      • POST databases/{tkn-db}/_replication creates a replication using the {rep-id} specified in the body or if none specified, a unique UUID.

      • PUT databases/{tkn-db}/_replication/{rep-id} upserts replication {rep-id}.

    • ENTERPRISE EDITION only:

      • Replications are distributed even across all available Sync Gateway nodes and so are not guaranteed to run on their originating node.

      • If a multi-node Sync Gateway cluster loses a subset of sync gateway nodes, the remaining nodes continue replication uninterrupted IF they have been configured to handle the replication (continuous and one-shot replications).

    Replication Definition

    All replications are 'initialized' by a replication definitionglossary icon in the configuration file or Admin REST API and operate within the context of a local database.

    • Configured replications use the database.{db-name}.replications property to add a replication definition to a local database.

    • REST API replications specify the local database and replication identity in the API POST/PUT request. Providing the replication definition parameters in the request body as a JSON string.

    Both scenarios are covered in Example 2. It summarizes the replication definitionglossary icon elements[2], which are covered in more detail in Database Configuration.

    Database-level Settings

    A number of database-level options are also especially relevant to Inter-Sync Gateway Replication, including:

    • sgreplicate_enabled — use this ENTERPRISE EDITION setting to allow the database to participate in Inter-Sync Gateway Replications.

    • database.delta_sync — use this setting to enable delta-sync replication on the database, it must be set if you want to use delta-sync in your replication definition.

    • sgreplicate_websocket_heartbeat_secs — use this setting to override the default (5 minute) heartbeat interval for websocket ping frames for this database.

    • database.sync — use this setting to specify the sync function logic — this is an essential part of access-control.

    • unsupported.sgr_tls_skip_verify — use this unsupported option to make development an testing easier by skipping verification of TLS certificates.

    Replication-level Settings

    Example 2. Replication Definition
    • Summary of Parameters

    • Configured Example

    • REST API Example

    This table summarize all the available configurable items.

    Data schema for the replication model

    Name Description Schema

    adhoc
    optional

    " About

    Use the Admin REST API’s adhoc parameter to specify that a replication is ad hoc rather than persistent.

    Behavior

    Ad hoc replications behave the same as normal replications, but they are automatically removed when their status changes to stopped. This will usually be on completion, but may also be as a result of user action.

    Constraints

    This parameter is NOT available to configured replications; only those initialized using the Admin REST API."
    Default : false

    boolean

    batch_size
    optional

    About

    Use the optional batch_size property to specify the number of changes to be included in a single batch during replication.

    integer

    cancel
    optional

    About

    Use this parameter on,y when you want to want to cancel an existing active replication.

    Constraints

    * This parameter is NOT available in configured replications; only those initialized using the Admin REST API.

    * NOTE that the body of the request must be the same as the replication’s replication definition for the cancellation request to be honoured. For example, if you requested continuous replication, the cancellation request must also contain the continuous field.
    Default : false

    boolean

    conflict_resolution_type
    optional

    About

    The conflict_resolution_type property defines the conflict resolution policy that Sync Gateway applies when resolving conflicting revisions.

    The default behavior is that automatic conflict resolution policy is applied.

    Valid options - default - localWins - remoteWins - custom

    Behavior

    * default - Selecting default applies the following conflict resolution policy * Deletes always win (the delete with longest revision history wins if both revisions are deletes) * The revision with the longest revision history wins (so, the one with most changes and consequently the highest revision Id). * localWins - Selecting localWins will result in local revisions always being the winner in any conflict.

    * remoteWins - Selecting remoteWins will result in remote revisions always being the winner in any conflict. * custom - Selecting custom specifies that you want to handle conflict resolution with your own application logic. You must provide this logic as a Javascript function by specifying it in using the custom-conflict-resolver parameter.

    Example

    ---- "conflict_resolution_type":"remoteWins" ----

    Constraints

    * Replications created prior to version 2.8 will default to default.
    Default : "default"

    string

    continuous
    optional

    About

    The continuous property specifies whether this replication will run in continuous mode.

    Behavior

    * continuous=true– In continuous mode, changes are immediately synced in accordance with the replication definition. * continuous=false– Detected changes are synced in accordance with the replication definition. The replication ceases once all revisions are processed.

    Constraints

    * Optional for stops and removes
    Default : false

    boolean

    custom_conflict_resolver
    optional

    About

    The optional custom_conflict_resolver property specifies the Javascript function that will be used to resolve conflicts, if the custom conflict resolution type is specified in the conflict_resolution_type.

    Options

    The property is mandatory when conflict_resolution_type=custom and will be ignored in all other cases.

    Using

    Provide the required logic in a Javascript function, as a string within backticks (see also the description for the sync function`.

    The function takes one parameter struct representing the conflict and comprising - the document id - the local document - the remote document

    The function returns a document struct representing the winning revision.

    Example

    ---- "custom_conflict_resolver":` function(conflict) { console.log("full remoteDoc doc: "+JSON.stringify(conflict.RemoteDocument)); return conflict.RemoteDocument; }` ----

    Constraints

    Using complex custom_conflict_resolver functions can noticeably degrade performance. Use a built-in resolver whenever possible.
    Default : "none"

    string

    direction
    optional

    About

    The mandatory direction property specifies whether the replication is push, pull or pushAndPull relative to this node.

    The property value is referenced by the remote property.

    Behavior

    * pull - changes are pulled from the remote database * push - changes are pushed to the remote database * pushAndPull - changes are both pushed-to and pulled-from the remote database

    Constraints

    Replications created prior to version 2.8 derive their direction from the source/target url of the replication.

    string

    enable_delta_sync
    optional

    About

    The optional enable_delta_sync parameter turns on delta sync for a replication. It works in conjunction with the database level setting delta_sync.enabled.

    Options

    * "enable_delta_sync": true, the replication can use delta sync (depending on delta_sync.enabled setting) * "enable_delta_sync": false, the replication cannot use delta sync

    Behavior

    The optional enable_delta_sync parameter works in conjunction with the database level delta_sync.enabled setting, to determine whether this replication uses delta sync.

    * If "delta_sync.enabled": true for both databases involved in the replication, then this parameter enables or disables its use for this specific replication. * In all other cases it has no effect and the replication runs without delta-sync.

    Constraints

    * Applies ONLY to Enterprise Edition deployments. * Depends upon the setting of the database level parameter delta_sync.enabled * Replications created prior to version 2.8 must run with "enable_delta_sync": false * Push replications will not use Delta Sync when pushing to a pre-2.8 target
    Default : false

    boolean

    filter
    optional

    About

    Use the optional filter`property to defines the function to be used to filter documents.

    Options

    A common value used when replicating from Sync Gateway is `sync_gateway/bychannel. This option limits the pull replication to a specific set of channels. You can specify the required channels using query_params.

    Behavior

    Works in conjunction with query_params to control the documents processed by the replication.

    Example

    ---- "filter":"sync_gateway/bychannel" ----

    Constraints

    OPTIONAL for stops and removes (even if defined during creation)

    string

    initial_state
    optional

    About

    The optional initial_state property is used to specify that the replication must be launched in 'Stopped' mode

    Behavior

    All replications are configured to start on Sync Gateway launch. So, if omitted, the state defaults to 'Running'.

    Constraints*

    Replications created prior to version 2.8 will all default to a state of 'Running'.
    Default** : "Running"

    string

    max_backoff_time
    optional

    The *max_backoff_time*property specifies the time-period (in minutes) during which Sync Gateway will attempt to reconnect lost or unreachable remote targets.

    On disconnection, Sync Gateway will do an exponential backoff up to the specified value, after which it will attempt to reconnect indefinitely every max_backoff_time minutes.

    If a zero value is specified, then Sync Gateway will do an exponential backoff up to an interval of five minutes before stopping the replication.

    NOTE - this value defaults to five minutes for replications created prior to version 2.8.

    integer

    password
    optional

    About

    Use password to provide the login password value for the accredited user running this replication.

    Behavior

    These details are used to authenticate credentials and approve access to data.

    Once provided and recorded, the password data is redacted and will not be displayed in either the configuration file or Admin REST API. A string of * will be displayed in its place.
    Default
    * : "mandatory"

    string

    purge_on_removal
    optional

    About

    The optional purge_on_removal property specifies, per replication, whether the removal of a channel triggers a purge.

    Options - true or false - Default = false - Document removals are ignored by receiving end

    Behavior

    If purge_on_removal=false, then the removal of channels is ignored (not purged) by the receiving end.

    Constraints

    * Applies only to PULL replications, including the PULL portion of a PUSHANDPULL replication.

    * Replications created prior to version 2.8 must be run with purge_on_removal=false.
    Default : false

    boolean

    query_params
    optional

    About

    The query_params property defines a set of key/value pairs used in the query string of the replication.

    Behavior

    This property works in conjunction with filters and channels to provide routing.

    Using

    You can use query_params’ channels function to pull from a specific set of `channels. To do so, you would also need to set the filter to sync_gateway/bychannels.

    Example

    [source,json] ---- "filter":"sync_gateway/bychannel", "query_params": { "channels":["channel.user1"] }, ----

    Constraints

    OPTIONAL for stops and removes (even if defined during creation)

    < string > array

    remote
    optional

    About

    The remote property represents the endpoint of a database for the remote Sync Gateway. That is, it identifies the remote Sync Gateway database that is the subject of this replication’s push, pull or pushAndPull action.

    Typically the endpoint will include URI, Port and Database name elements.

    You can also include user credentials in the URL, in the form <username>:<password>. The credentials relate to an existing Sync Gateway user on the remote server.

    Example `"remote": "http://user:password@example.com:4985/db1-remote"; `

    Format

    * a string containing a valid URL for a (remote) Sync Gateway database. * an object whose url property contains the Sync Gateway database URL.

    Behavior

    Dependent upon setting of direction.

    If direction is : - pull, 'remote' defines the remote cluster from which data is pulled - push, 'remote' defines the remote cluster to which data is pushed - pushAndPull, 'remote' defines the push configuration.

    Example

    [source,json] ---- "remote": "http://www.example.com:4984/sample-database", ----

    string

    replication_id
    optional

    About

    The replication_id property specifies either:

    * For NEW replications, the ID to be assigned to the the replication. If no replication_id is specified, Sync Gateway will assign a random UUID to new replications.

    * For existing replications, this is the ID of the required replication.

    * If cancel=true, this is the id of the active replication task to be cancelled.

    Constraints

    If this is specified in the body of a POST or PUT request then it must be the same value as specified in the request URL.

    string

    username
    optional

    About

    Use username to provide the name of the accredited user running this replication.

    Behavior

    These details are used to authenticate credentials and approve access to data

    Once provided and recorded, the username data is redacted and will not be displayed in either the configuration file or Admin REST API. A string of * will be displayed in its place.
    Default
    * : "Mandatory"

    string

    This is an example of a replication definition. Its purpose is to illustrate all configurable items in use and so should not be considered a working example.

    It creates a replication with the replication_ID of db1-rep-id1-pull-oneshot on a local database db1-local_, pulling data from a remote database db1-remote.

    "databases": {
      " db1": {                                                (1)
        "bucket":"db1",
        "server": "couchbase://cb-server",
        // ... other DB config ..
        "sgreplicate_enabled": true,                          (2)
        "replications":
        "db1-rep-id1-pull-oneshot":                         (3)
        "direction": "pull",                              (4)
        "remote": "https://example.com:4984/remote_db1",
        "user": "user1",                                  (5)
        "password": "password",
        "batch_size": 1000,                               (6)
        "conflict_resolution_type": "custom",             (7)
        "custom_conflict_resolver": "",                   (8)
        "continuous": false,                              (9)
        "enable_delta_sync": false,                       (10)
        "filter": "sync_gateway/bychannel",               (11)
        "query_params": ["channel.user1"]                 (12)
        "max_backoff_time": 5,                            (13)
        "purge_on_removal": false                         (14)
        "state": "running"                                (15)
      }
    }
    1 All replications are defined at database level within the context of a local database (e.g. DB1)
    2 Opt in to replication
    3 Define the replication_id
    4 Pull changes from the remote database at the specified url.
    5 Authenticate with the provided credentials. This user must have read and write access to both the local and remote databases.
    6 Batch together up to 1000 revisions at a time. This improve replication performance but consumes more memory resources.
    7 Apply a custom conflict resolution policy.
    8 Provide a working Javascript function to apply the required resolution policy.
    9 By setting continuous=false, we are creating a one-shot replication. We could also have omitted this parameter as it defaults to false.
    10 Don’t use delta-sync; the default behavior.
    11 Filter documents by channel.
    12 Replicate only those documents tagged with the channel names "user1".
    13 Wait no more than 5 minutes between retries after network failure; default behavior.
    14 Don’t purge following removal of a channel; the default behavior.
    15 Start the replicator immediately and on Sync Gateway node re(start);. We could also have omitted this parameter as this is the default behavior.

    This is an example of a replication definition as you might submit it to the Admin REST API.using curl. Its purpose is to illustrate all configurable items in use and so should not be considered a working example.

    It creates a replication with the replication_ID of db1-rep-id1-pull-oneshot on a local database db1-local_, pulling data from a remote database db1-remote.

    curl --location --request POST 'http://localhost:4985/db1-local/_replication/db1-rep-id1-pull-oneshot' \ (1)
    --header 'Content-Type: application/json' \
    --dataraw '{
    "replication_id": "db1-rep-id1-pull-oneshot" (2)
        "direction": "pull",                              (3)
        "remote": "https://example.com:4984/remote_db1",
        "user": "user1",                                  (4)
        "password": "password",
        "batch_size": 1000,                               (5)
        "conflict_resolution_type": "custom",             (6)
        "custom_conflict_resolver": "",                   (7)
        "continuous": false,                              (8)
        "enable_delta_sync": false,                       (9)
        "filter": "sync_gateway/bychannel",               (10)
        "query_params": ["channel.user1"]                 (11)
        "max_backoff_time": 5,                            (12)
        "purge_on_removal": false                         (13)
        "state": "running"                                (14)
      "adhoc": false, (15)
      "cancel": false (16)
    }'
    1 All replications take place at database level and in the context of a local database. Here we are setting the replication in the context of db1-local
    2 Define the replication_id
    3 Pull changes from the remote database at the specified url.
    4 Authenticate with the provided credentials. This user must have read and write access to both the local and remote databases.
    5 Batch together up to 1000 revisions at a time. This improve replication performance but consumes more memory resources.
    6 Apply a custom conflict resolution policy.
    7 Provide a working Javascript function to apply the required resolution policy.
    8 By setting continuous=false, we are creating a one-shot replication. We could also have omitted this parameter as it defaults to false.
    9 Don’t use delta-sync; the default behavior.
    10 Filter documents by channel.
    11 Replicate only those documents tagged with the channel names "user1".
    12 Wait no more than 5 minutes between retries after network failure; default behavior.
    13 Don’t purge following removal of a channel; the default behavior.
    14 Start the replicator immediately and on Sync Gateway node re(start);. We could also have omitted this parameter as this is the default behavior.
    15 Setting adhoc=false marks this as a persistent replication. The definition will survive Sync Gateway node restarts. This the default behaviour if this parameter is omitted.+
    16 Set cancel=true to cancel an initialized replication; otherwise you can omit this parameter.

    Generic Constraints

    Replication

    All active nodes in an active cluster must be running Sync Gateway version 2.8+.

    ENTERPRISE EDITION

    All replications are distributed evenly across available nodes. This means they cannot be guaranteed to run on the node from which they originate.

    Access rights

    The user running the replication must have read and write access to the data being replicated. This is not enforced by the system. Use your sync function to ensure a consistent approach is applied across all clusters.

    Mixing Inter-Sync Gateway Replication Versions

    Versions of inter-Sync Gateway replications pre- and post-2.8 can legitimately be in use at the same time, especially during transition. However, you should avoid initializing identical pre-2.8 (SG Replicate) and 2.8+ replications.

    Running Configured Replications

    Replications in the configuration file start automatically whenever Sync Gateway is (re)started. Unless you inhibit this by adding an "initial_state": "stopped" parameter to the replication definition — see: initial_state. You can manually start 'stopped' replication using Starting a replication.

    Example 3. Configured Replications — Continuous and One-shot
    • Continuous

    • One-shot

    //  . . . other configuration entries
    "db1-rep-id1-pull-cont":
    "replication_id": "db1-rep-id1-pull-cont",
    "direction": "pull",
    "continuous": true (1)
    "purge-on-removal": true,
    "remote": "http://user:password@example.com:4985/db1-remote",
    "filter":"sync_gateway/bychannel",
    "query_params": {
      "channels": ["channel1.user1"]
    }
    //  . . . other configuration entries
    1 Make this a continuous replication that remains running, listening for changes to process. Because it is also persistent, it will start automatically following Sync Gateway node restarts (state defaults to running).
    2 The remote URL can also include the credentials for an existing Sync Gateway user on the remote server.
    //. . . other configuration entries
    "db1-rep-id3-pull-oneshot":
    "replication_id": "db1-rep-id3-pull-oneshot", (1)
    "direction": "pull",
    "remote": "http://user1:password@example.com:4985/db1-remote", (2)
    "filter": "sync_gateway/bychannel",
    "query_params": { "channels": ["channel1.user1"] }
    //  . . . other configuration entries
    1 This a a one-shot replication because the continuous parameter defaults to false.
    2 The remote URL can also include the credentials for an existing Sync Gateway user on the remote server.

    Running Admin REST API Replications

    Replications initialized by sending a POST, or PUT, request to the _replication endpoint will start running automatically, unless the "initial_state": "stopped" parameter is specified. with a JSON object defining the replication parameters — as shown in Example 4.

    You can run multiple replications simultaneously with different replication topologies, provided both databases being synchronized have the same sync function.

    You can submit requests using the curl utility (as in these examples) or an application such as Postman.

    Example 4. Submitting API Requests
    • Continuous Pull Replication

    • One-shot

    • Ad-hoc

    This example initializes a persistent, continuous, replication between a local database and one on a remote Sync Gateway instance.

    curl --location --request POST 'http://localhost:4985/db1-local/_replication/' \
    --header 'Content-Type: application/json' \
    --dataraw '{
    "replication_id": "db1-rep-id1-pull-cont",
    "direction": "pull",
    "continuous": true (1)
    "purge-on-removal": true,
    "remote": "http://user:password@example.com:4985/db1-remote",
    "filter":"sync_gateway/bychannel",
    "query_params": {
      "channels": ["channel1.user1"]
    }
    }'

    This example initializes a persistent, one-shot, replication between a local database and one on a remote Sync Gateway instance.

    The replication will run once, after a short delay to allow the Rest API to start. It will then run once after each Sync Gateway restart and-or when manually initiated using the _replicationStatus endpoint — see Inter Sync Gateway Sync - Manage.

    curl --location --request POST 'http://localhost:4985/db1-local/_replication/' \
    --header 'Content-Type: application/json' \
    --dataraw '{
    "replication_id": "db1-rep-id3-pull-oneshot", (1)
    "direction": "pull",
    "remote": "http://user1:password@example.com:4985/db1-remote", (2)
    "filter": "sync_gateway/bychannel",
    "query_params": { "channels": ["channel1.user1"] }
    }'
    1 This a a one-shot replication because the continuous parameter defaults to false.
    2 The remote URL can also include the credentials for an existing Sync Gateway user on the remote server.
    curl --location --request POST 'http://localhost:4985/db1-local/_replication/' \
    --header 'Content-Type: application/json' \
    --dataraw '{
    "replication_id": "db1-rep-id1-pull-adhoc",
    "adhoc": true, (1)
    "direction": "pull",
    "purge-on-removal": true,
    "remote": "http://user:password@example.com:4985/db1-remote",
    "filter":"sync_gateway/bychannel",
    "query_params": {
      "channels": ["channel1.user1"]
    }
    }'
    1 Run this replication as an ad hoc one. It will run once only, process all changes but not survive Sync Gateway restarts


    1. This parameter is not available in the configuration file.
    2. The definitions apply to configured and API replications).