-
Notifications
You must be signed in to change notification settings - Fork 144
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: iroh-gossip #1149
feat: iroh-gossip #1149
Conversation
00f5539
to
0129484
Compare
608c72a
to
119d6b9
Compare
@@ -0,0 +1,182 @@ | |||
use rand::{ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
missing unit tests, ideally some proptests for the pieces in this file
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added basic unit tests for IndexSet
, TimerMap
and the new base32
utils.
What parts would you want to proptest? Can do, just not super sure if/where it would make sense here.
## Description This PR adds `iroh-sync`, a document synchronization protocol, to iroh, and integrates with `iroh-net`, `iroh-gossip` and `iroh-bytes`. * At the core is the `iroh-sync` crate, with a set reconciliation algorithm implemented by @dignifiedquire. See [the old iroh-sync repo](https://github.com/n0-computer/iroh-sync/) for the prehistory and #1216 for the initial PR (fully included in this PR, and by now outdated) * Iroh sync is integrated in the iroh node, with iroh-gossip, in the RPC interface, and the CLI. * `LiveSync` is the handle to an actor that integrates sync with [gossip](#1149 ) to broadcast and receive document updates from peers. For each open document a gossip swarm is joined with a `TopicId` derived from the doc namespace. * mod `download` contains the new downloader. It will be improved in #1344 . * mod `client` is the new high-level RPC client. It currently only has methods for dealing with docs and sync, other things should be added once we merged this. CLI commands for sync are in `commands/sync.rs`. Will be much better with #1356 . * `examples/sync.rs` has a REPL to modify and sync docs. It does a full setup without using the iroh console. Also includes code to sync directories, and a hammer command for load testing. * The PR also introduces `iroh::client::Iroh`, a wrapper around the RPC client, and `iroh::client::Doc`, a wrapper around RPC client for a single document ## Notes & open questions #### Should likely happen before merge: * [x] Make `iroh_sync::Store:::list_authors` and `list_replicas` return iterators `iroh-sync` *fixed in #1366 * * [ ] Add `iroh_sync::Store::close_replica` * [x] `ContentStatus` in `on_insert` callback is reported as `Ready` if the content is still `baomap::PartialEntry` (in-process download) *fixed in a8e8093* #### Can happen after merge, but before `0.6` release * [ ] Implement `AuthorImport` and `AuthorShare` RPC & CLI commands * [ ] sync store `list_namespaces` and `list_authors` internally collect, return iterator instead * [ ] Fix cross-compiles to arm/android. See cross-rs/cross#1311 * [ ] Ensure that fingerpring calculation is efficient and/or cached for large documents. Currently calculating the initial fingerprint iterates once over all entries in a document. * [ ] Make content downloads be more reliable * [ ] Add some way to download content from peers independent of the first insertion event for a remote entry. The downloader with retries is tracked in #1334 and 1344, but independent of that, we still would currently only ever try to queue a download when the `on_insert` callback triggers, which is only once. There should be a way, even if manual for now, to try to download missing content in a replica from peers. * [ ] during `iroh-sync` sync include info if content is available for each entry * [ ] Add basic peer management and persistence. Currently live sync will stop to do anything after a node restart. * [ ] Persist the addressbook of peers for a document, to reconnect when restarting the node * [ ] Implement `PeerAdd` and `PeerList` RPC & CLI commands. The latter needs changes in `iroh-net` to expose information of currently-connected peers and their peer info. * [ ] Make read-only replicas possible * [ ] Improve reliablity of replica sync. * sync is triggered on each `NeighborUp` event from gossip. check that we don't sync too much. * maybe include peer info in gossip messages, to queue syncs with those (but not all at once) * track and exchange the timestamp of last full sync for peers, to know if you missed gossiped message and react accordingly * add more tests with peers coming and leaving #### Open questions * [ ] `iroh_sync::EntrySignature` should the signatures include a namespace prefix? * [ ] do we want the 1:1 mapping of `NamespaceId`and gossip `TopicId`, or would the topic id as a hash be better? #### Other TODOs collected from the code * [ ] Port `hammer` and `fs` commands from REPL example to iroh cli * [ ] docs/naming: settle terminology about keypairs, private/secret/signing keys, public keys/identifiers and make docs and symbols consistent * [ ] Make `bytes_get` streaming in the RPC interface * [ ] Allow to unset the subscription on a replica * [ ] `iroh-sync` shouldn't depend on `iroh-bytes` only for `Hash` type -> #1354 * [ ] * [ ] Move `sync::live::PeerSource` to iroh-net or even better -> #1354 * [ ] `StoreInstance::put` propagate error and verify timestamp is reasonable. * [ ] `StoreInstance::get_range` implement inverted range * [ ] `iroh_sync`: Remove some items only used in tests (marked with #[cfg(test)]) * [ ] `iroh_sync` fs store: verify get method fetches all keys with this namespace * [ ] `ranger::SimpleStore::get_range`: optimize * [ ] `ranger::Peer` avoid allocs? * [ ] `fs::StoreInstance::get_fingerprint` optimize * [ ] `SyncEngine::doc_subscribe` remove unwrap, handle error ## Change checklist - [x] Self-review. - [x] Documentation updates if relevant. - [ ] Tests if relevant. --------- Co-authored-by: dignifiedquire <me@dignifiedquire.com> Co-authored-by: Asmir Avdicevic <asmir.avdicevic64@gmail.com> Co-authored-by: Kasey <klhuizinga@gmail.com>
Description
This PR adds
iroh-gossip
. See n0-computer/iroh-sync#1 for the prehistory.iroh-gossip exposes a pubsub-like API to broadcast messages on topics. Peers join a topic by connecting to at least one other peer. They can then broadcast messages to all other peers who joined a topic, and subscribe to messages. The protocol is based on HyParView, which gives us an overlay network of all nodes that joined a topic; and PlumTree to create an epidemic broadcast tree for reliably broadcasting messages to all online peers.
The new crate is structured as follows:
The protocol implementation (mod
proto
) as an IO-less state machine. It exposes a single method to handleInEvent
s and returns an iterator ofOutEvent
s which must be handled by the implementor (register timers, send messages, emit events). The implementation is generic over aPeerAddress
trait, which should be an identifier for a peer. In our case, it will be aPeerId
.TopicId
, a 32 byte identifier. Topics are seperate swarms and broadcast scopes. The HyParView and PlumTree algorithms both work in the scope of a single topic. Thus, joining multiple topics increases the number of open connections to peers and the size of the local routing table.state::State
. It mostly dispatches messages based on the topic ID into the matchingtopic::State
. This is the actual protocol state for a topic. It, again, mostly dispatches messages onto the actual protocols:hyparview::State
andplumtree::State
.The
net
module provides networking for the IO-less protocol impl, through theMagicEndpoint
from iroh-net, with a (I think) quite straightforward actor implementation.The
net
module has a smoke test, and the PR also includes achat
example that adds signed messages on the trustless base layer.What this mostly needs is, I think, more testing and validation of the protocol implementation. For this to be practical, I think we should formulate some base usecase scenarios.
The
proto
module has a very simple simulator and some tests that can take environment variables to increase the cluster size. They passed with 10.000 nodes. They also report some stats. The numbers look OK-ish, but I'd love to compare them properly with either other implementations or estimations from the original paper or other resoures. However, those are hard to find.Notes & open questions
plumtree
module builds an unbounded cache of all broadcast messages. This will have to change. The simplest might be to move to an LRU cache for now. In the future, we might want to expose a method for the module to request expired messages from a backing storage or so. (The module needs to have a cache of messages to reply for requests by other peers. However these will only come within a reasonable amount of time after message publication / reception.) this is now implemented with an expiring cacheChange checklist