ssb-ebt

Replicates SSB feeds using the efficient "Epidemic broadcast tree" algorithm.

This module is an adapter for the module epidemic-broadcast-trees to work with secure-scuttlebutt. Supersedes ssb-replicate.

Installation

Prerequisites:

• Requires Node.js 10 or higher
• Requires ssb-db or ssb-db2 version 5.0 or higher

npm install --save ssb-ebt

Add this secret-stack plugin like this:

 const SecretStack = require('secret-stack')
 const caps = require('ssb-caps')

 const createSsbServer = SecretStack({ caps })
     .use(require('ssb-master'))
     .use(require('ssb-db'))
+    .use(require('ssb-ebt'))
     .use(require('ssb-friends'))
     // ...

Usage

ssb-ebt itself does NOT trigger replication automatically after it's installed, instead, you need to call its API methods yourself (primarily request or block), or use a scheduler module such as ssb-replication-scheduler.

You should really install ssb-replication-scheduler, but if you have a strong reason not to, this is how you should use ssb-ebt:

• You need to call ssb.ebt.request() for all feeds that you want AND for all the feeds that you allow other peers to fetch from you
• In particular, this means you need to call ssb.ebt.request(ssb.id, true) (i.e. "replicate myself") if you want other peers to replicate you
• Replicating yourself is also important for recovering your feed from zero
• Note that ssb.ebt.block takes a destination AND origin, while ssb.ebt.request only takes a destination because the origin is assumed to be yourself
• You need to call ssb.ebt.block(ssb.id, feedId, true) for all feedId that you (ssb.id) want to prevent replication, but you also have to call ssb.ebt.block(alice, bob, true) for all cases where alice blocks bob. This is important to respect Alice's wish for her data to not be shared with Bob.

ssb.ebt.request(destination, replicating, formatName) ("sync" muxrpc API)

Request that the SSB feed ID destination be replicated. replication is a boolean, where true indicates we want to replicate the destination. If set to false, replication is stopped. formatName is optional and used to specify the specific EBT instance, otherwise the first where isFeed is true for destination is used.

Returns undefined, always.

ssb.ebt.block(origin, destination, blocking) ("sync" muxrpc API)

Computes that origid does not want to replicate destination's feed. Also disallows other peers (who have this same ssb-ebt installed) to pass on data to them.

origin is the SSB feed ID of the peer who created the block, destination is the SSB feed ID of the peer being blocked, and blocking is a boolean that indicates whether to enable the block (true) or to unblock (false).

formatName is optional and used to specify the specific EBT instance, otherwise the first where isFeed is true for origin is used.

Returns undefined, always.

ssb.ebt.forget(destination) ("sync" muxrpc API)

Same as ssb.ebt.request(destination, false), but also cleans up any persistent state used by the EBT instance for this given destination ID.

Returns undefined, always.

ssb.ebt.peerStatus(id) ("sync" muxrpc API)

Query the status of replication for a given SSB feed ID id. Returns a JSON object showing the replication state for all peers we are currently connected to.

The output looks like this:

ssb.ebt.registerFormat(methods) ("sync" muxrpc API)

Register a new format for replication. Note this does not have to be a new feed format, it could also be indexed replication or sliced replication. See formats folder for examples.

By registering a format you create a new EBT instance used for replicating feeds using that format. This means its own clock. Message will be replicated using the replicateFormat API. The methods argument must implement the following functions. The example below shows the implementation for 'classic' ed25519 SSB feeds.

ssb.ebt.setClockForSlicedReplication(feedId, sequence, formatName) ("sync" muxrpc API)

Sets the internal clock of a feed to a specific sequence. Note this does not start replicating the feed, it only updates the clock. By combining this with clock it is possible do to sliced replication with a remote peer where say only the latest 100 messages of a feed is replicated.

(Internal) ssb.ebt.replicate(opts) ("duplex" muxrpc API)

Creates a duplex replication stream to the remote peer. When two peers connect, the peer who initiated the call (the client) should call this. You do not need to call this method, it is called automatically in ssb-ebt whenever our peer connects to a remote peer. opts is an object with one field: version.

(Internal) ssb.ebt.replicateFormat(opts) ("duplex" muxrpc API)

Creates a duplex replication stream to the remote peer. This behaves similar to replicate except it takes an extra field format specifying what is transferred over this EBT stream. Classic feeds are still replicated using replicate while this will be used to replicate other feed formats.

(Internal) ssb.ebt.clock(opts, cb) ("async" muxrpc API)

Gets the current vector clock of a remote peer. opts is an object with one field: format specifying what format to get the vector clock for. Defaults to 'classic'.

Testing and debugging

There are several scripts in ./debug which can be used for testing EBT replication.

Use ./debug/remote.js <address> to connect to an SSB peer running EBT. Running this won't store anything locally, it will just download everything and drop it on the floor. This is used to test performance of EBT on a server.

We normally see values between 2k and 3k messages per second, in other words, replicates 100k messages in under a minute.

License

MIT