On heavily crowded networks, we are seeing many attempted connections to our node every second.
Often these connections come from peers that have just been disconnected. This can be for a number of reasons including:
- We have deemed them to be not as useful as other peers
- They have performed poorly
- They have dropped the connection with us
- The connection was spontaneously lost
- They were randomly removed because we have too many peers
In all of these cases, if we have reached or exceeded our target peer limit, there is no desire to accept new connections immediately after the disconnect from these peers. In fact, it often costs us resources to handle the established connections and defeats some of the logic of dropping them in the first place.
This PR adds a timeout, that prevents recently disconnected peers from reconnecting to us.
Technically we implement a ban at the swarm layer to prevent immediate re connections for at least 10 minutes. I decided to keep this light, and use a time-based LRUCache which only gets updated during the peer manager heartbeat to prevent added stress of polling a delay map for what could be a large number of peers.
This cache is bounded in time. An extra space bound could be added should people consider this a risk.
Co-authored-by: Diva M <divma@protonmail.com>
## Issue Addressed
Adds self rate limiting options, mainly with the idea to comply with peer's rate limits in small testnets
## Proposed Changes
Add a hidden flag `self-limiter` this can take no value, or customs values to configure quotas per protocol
## Additional Info
### How to use
`--self-limiter` will turn on the self rate limiter applying the same params we apply to inbound requests (requests from other peers)
`--self-limiter "beacon_blocks_by_range:64/1"` will turn on the self rate limiter for ALL protocols, but change the quota for bbrange to 64 requested blocks per 1 second.
`--self-limiter "beacon_blocks_by_range:64/1;ping:1/10"` same as previous one, changing the quota for ping as well.
### Caveats
- The rate limiter is either on or off for all protocols. I added the custom values to be able to change the quotas per protocol so that some protocols can be given extremely loose or tight quotas. I think this should satisfy every need even if we can't technically turn off rate limits per protocol.
- This reuses the rate limiter struct for the inbound requests so there is this ugly part of the code in which we need to deal with the inbound only protocols (light client stuff) if this becomes too ugly as we add lc protocols, we might want to split the rate limiters. I've checked this and looks doable with const generics to avoid so much code duplication
### Knowing if this is on
```
Feb 06 21:12:05.493 DEBG Using self rate limiting params config: OutboundRateLimiterConfig { ping: 2/10s, metadata: 1/15s, status: 5/15s, goodbye: 1/10s, blocks_by_range: 1024/10s, blocks_by_root: 128/10s }, service: libp2p_rpc, service: libp2p
```
I've needed to do this work in order to do some episub testing.
This version of libp2p has not yet been released, so this is left as a draft for when we wish to update.
Co-authored-by: Diva M <divma@protonmail.com>
Our custom RPC implementation is lagging from the libp2p v50 version.
We are going to need to change a bunch of function names and would be nice to have consistent ordering of function names inside the handlers.
This is a precursor to the libp2p upgrade to minimize merge conflicts in function ordering.
- there was a bug in responding range blob requests where we would incorrectly label the first slot of an epoch as a non-skipped slot if it were skipped. this bug did not exist in the code for responding to block range request because the logic error was mitigated by defensive coding elsewhere
- there was a bug where a block received during range sync without a corresponding blob (and vice versa) was incorrectly interpreted as a stream termination
- RPC size limit fixes.
- Our blob cache was dead locking so I removed use of it for now.
- Because of our change in finalized sync batch size from 2 to 1 and our transition to using exact epoch boundaries for batches (rather than one slot past the epoch boundary), we need to sync finalized sync to 2 epochs + 1 slot past our peer's finalized slot in order to finalize the chain locally.
- use fork context bytes in rpc methods on both the server and client side
## Issue Addressed
Implementing the light_client_gossip topics but I'm not there yet.
Which issue # does this PR address?
Partially #3651
## Proposed Changes
Add light client gossip topics.
Please list or describe the changes introduced by this PR.
I'm going to Implement light_client_finality_update and light_client_optimistic_update gossip topics. Currently I've attempted the former and I'm seeking feedback.
## Additional Info
I've only implemented the light_client_finality_update topic because I wanted to make sure I was on the correct path. Also checking that the gossiped LightClientFinalityUpdate is the same as the locally constructed one is not implemented because caching the updates will make this much easier. Could someone give me some feedback on this please?
Please provide any additional information. For example, future considerations
or information useful for reviewers.
Co-authored-by: GeemoCandama <104614073+GeemoCandama@users.noreply.github.com>
## Issue Addressed
#3704
## Proposed Changes
Adds is_syncing_finalized: bool parameter for block verification functions. Sets the payload_verification_status to Optimistic if is_syncing_finalized is true. Uses SyncState in NetworkGlobals in BeaconProcessor to retrieve the syncing status.
## Additional Info
I could implement FinalizedSignatureVerifiedBlock if you think it would be nicer.
## Issue Addressed
Partially addresses #3651
## Proposed Changes
Adds server-side support for light_client_bootstrap_v1 topic
## Additional Info
This PR, creates each time a bootstrap without using cache, I do not know how necessary a cache is in this case as this topic is not supposed to be called frequently and IMHO we can just prevent abuse by using the limiter, but let me know what you think or if there is any caveat to this, or if it is necessary only for the sake of good practice.
Co-authored-by: Pawan Dhananjay <pawandhananjay@gmail.com>
## Proposed Changes
With a few different changes to the gossip topics in flight (light clients, Capella, 4844, etc) I think this simplification makes sense. I noticed it while plumbing through a new Capella topic.
This PR adds some health endpoints for the beacon node and the validator client.
Specifically it adds the endpoint:
`/lighthouse/ui/health`
These are not entirely stable yet. But provide a base for modification for our UI.
These also may have issues with various platforms and may need modification.
## Issue Addressed
New lints for rust 1.65
## Proposed Changes
Notable change is the identification or parameters that are only used in recursion
## Additional Info
na
## Issue Addressed
Updates discv5
Pending on
- [x] #3547
- [x] Alex upgrades his deps
## Proposed Changes
updates discv5 and the enr crate. The only relevant change would be some clear indications of ipv4 usage in lighthouse
## Additional Info
Functionally, this should be equivalent to the prev version.
As draft pending a discv5 release
* add capella gossip boiler plate
* get everything compiling
Co-authored-by: realbigsean <sean@sigmaprime.io
Co-authored-by: Mark Mackey <mark@sigmaprime.io>
* small cleanup
* small cleanup
* cargo fix + some test cleanup
* improve block production
* add fixme for potential panic
Co-authored-by: Mark Mackey <mark@sigmaprime.io>
## Issue Addressed
I missed this from https://github.com/sigp/lighthouse/pull/3491. peers were being banned at the behaviour level only. The identify errors are explained by this as well
## Proposed Changes
Add banning and unbanning
## Additional Info
Befor,e having tests that catch this was hard because the swarm was outside the behaviour. We could now have tests that prevent something like this in the future
## Issue Addressed
We currently subscribe to attestation subnets as soon as the subscription arrives (one epoch in advance), this makes it so that subscriptions for future slots are scheduled instead of done immediately.
## Proposed Changes
- Schedule subscriptions to subnets for future slots.
- Finish removing hashmap_delay, in favor of [delay_map](https://github.com/AgeManning/delay_map). This was the only remaining service to do this.
- Subscriptions for past slots are rejected, before we would subscribe for one slot.
- Add a new test for subscriptions that are not consecutive.
## Additional Info
This is also an effort in making the code easier to understand
## Issue Addressed
Resolves#3351
## Proposed Changes
Returns a `ResourceUnavailable` rpc error if we are unable to serve full payloads to blocks by root and range requests because the execution layer is not synced.
## Additional Info
This PR also changes the penalties such that a `ResourceUnavailable` error is only penalized if it is an outgoing request. If we are syncing and aren't getting full block responses, then we don't have use for the peer. However, this might not be true for the incoming request case. We let the peer decide in this case if we are still useful or if we should be banned.
cc @divagant-martian please let me know if i'm missing something here.
## Issue Addressed
N/A
## Proposed Changes
Since Rust 1.62, we can use `#[derive(Default)]` on enums. ✨https://blog.rust-lang.org/2022/06/30/Rust-1.62.0.html#default-enum-variants
There are no changes to functionality in this PR, just replaced the `Default` trait implementation with `#[derive(Default)]`.
## Overview
This rather extensive PR achieves two primary goals:
1. Uses the finalized/justified checkpoints of fork choice (FC), rather than that of the head state.
2. Refactors fork choice, block production and block processing to `async` functions.
Additionally, it achieves:
- Concurrent forkchoice updates to the EL and cache pruning after a new head is selected.
- Concurrent "block packing" (attestations, etc) and execution payload retrieval during block production.
- Concurrent per-block-processing and execution payload verification during block processing.
- The `Arc`-ification of `SignedBeaconBlock` during block processing (it's never mutated, so why not?):
- I had to do this to deal with sending blocks into spawned tasks.
- Previously we were cloning the beacon block at least 2 times during each block processing, these clones are either removed or turned into cheaper `Arc` clones.
- We were also `Box`-ing and un-`Box`-ing beacon blocks as they moved throughout the networking crate. This is not a big deal, but it's nice to avoid shifting things between the stack and heap.
- Avoids cloning *all the blocks* in *every chain segment* during sync.
- It also has the potential to clean up our code where we need to pass an *owned* block around so we can send it back in the case of an error (I didn't do much of this, my PR is already big enough 😅)
- The `BeaconChain::HeadSafetyStatus` struct was removed. It was an old relic from prior merge specs.
For motivation for this change, see https://github.com/sigp/lighthouse/pull/3244#issuecomment-1160963273
## Changes to `canonical_head` and `fork_choice`
Previously, the `BeaconChain` had two separate fields:
```
canonical_head: RwLock<Snapshot>,
fork_choice: RwLock<BeaconForkChoice>
```
Now, we have grouped these values under a single struct:
```
canonical_head: CanonicalHead {
cached_head: RwLock<Arc<Snapshot>>,
fork_choice: RwLock<BeaconForkChoice>
}
```
Apart from ergonomics, the only *actual* change here is wrapping the canonical head snapshot in an `Arc`. This means that we no longer need to hold the `cached_head` (`canonical_head`, in old terms) lock when we want to pull some values from it. This was done to avoid deadlock risks by preventing functions from acquiring (and holding) the `cached_head` and `fork_choice` locks simultaneously.
## Breaking Changes
### The `state` (root) field in the `finalized_checkpoint` SSE event
Consider the scenario where epoch `n` is just finalized, but `start_slot(n)` is skipped. There are two state roots we might in the `finalized_checkpoint` SSE event:
1. The state root of the finalized block, which is `get_block(finalized_checkpoint.root).state_root`.
4. The state root at slot of `start_slot(n)`, which would be the state from (1), but "skipped forward" through any skip slots.
Previously, Lighthouse would choose (2). However, we can see that when [Teku generates that event](de2b2801c8/data/beaconrestapi/src/main/java/tech/pegasys/teku/beaconrestapi/handlers/v1/events/EventSubscriptionManager.java (L171-L182)) it uses [`getStateRootFromBlockRoot`](de2b2801c8/data/provider/src/main/java/tech/pegasys/teku/api/ChainDataProvider.java (L336-L341)) which uses (1).
I have switched Lighthouse from (2) to (1). I think it's a somewhat arbitrary choice between the two, where (1) is easier to compute and is consistent with Teku.
## Notes for Reviewers
I've renamed `BeaconChain::fork_choice` to `BeaconChain::recompute_head`. Doing this helped ensure I broke all previous uses of fork choice and I also find it more descriptive. It describes an action and can't be confused with trying to get a reference to the `ForkChoice` struct.
I've changed the ordering of SSE events when a block is received. It used to be `[block, finalized, head]` and now it's `[block, head, finalized]`. It was easier this way and I don't think we were making any promises about SSE event ordering so it's not "breaking".
I've made it so fork choice will run when it's first constructed. I did this because I wanted to have a cached version of the last call to `get_head`. Ensuring `get_head` has been run *at least once* means that the cached values doesn't need to wrapped in an `Option`. This was fairly simple, it just involved passing a `slot` to the constructor so it knows *when* it's being run. When loading a fork choice from the store and a slot clock isn't handy I've just used the `slot` that was saved in the `fork_choice_store`. That seems like it would be a faithful representation of the slot when we saved it.
I added the `genesis_time: u64` to the `BeaconChain`. It's small, constant and nice to have around.
Since we're using FC for the fin/just checkpoints, we no longer get the `0x00..00` roots at genesis. You can see I had to remove a work-around in `ef-tests` here: b56be3bc2. I can't find any reason why this would be an issue, if anything I think it'll be better since the genesis-alias has caught us out a few times (0x00..00 isn't actually a real root). Edit: I did find a case where the `network` expected the 0x00..00 alias and patched it here: 3f26ac3e2.
You'll notice a lot of changes in tests. Generally, tests should be functionally equivalent. Here are the things creating the most diff-noise in tests:
- Changing tests to be `tokio::async` tests.
- Adding `.await` to fork choice, block processing and block production functions.
- Refactor of the `canonical_head` "API" provided by the `BeaconChain`. E.g., `chain.canonical_head.cached_head()` instead of `chain.canonical_head.read()`.
- Wrapping `SignedBeaconBlock` in an `Arc`.
- In the `beacon_chain/tests/block_verification`, we can't use the `lazy_static` `CHAIN_SEGMENT` variable anymore since it's generated with an async function. We just generate it in each test, not so efficient but hopefully insignificant.
I had to disable `rayon` concurrent tests in the `fork_choice` tests. This is because the use of `rayon` and `block_on` was causing a panic.
Co-authored-by: Mac L <mjladson@pm.me>
## Issue Addressed
Deprecates the step parameter in the blocks by range request
## Proposed Changes
- Modifies the BlocksByRangeRequest type to remove the step parameter and everywhere we took it into account before
- Adds a new type to still handle coding and decoding of requests that use the parameter
## Additional Info
I went with a deprecation over the type itself so that requests received outside `lighthouse_network` don't even need to deal with this parameter. After the deprecation period just removing the Old blocks by range request should be straightforward
## Issue Addressed
Which issue # does this PR address?
## Proposed Changes
Please list or describe the changes introduced by this PR.
## Additional Info
Please provide any additional information. For example, future considerations
or information useful for reviewers.
Code simplifications using `Option`/`Result` combinators to make pattern-matches a tad simpler.
Opinions on these loosely held, happy to adjust in review.
Tool-aided by [comby-rust](https://github.com/huitseeker/comby-rust).
## Issue Addressed
N/A
## Proposed Changes
Previously, we were using `Sleep::is_elapsed()` to check if the shutdown timeout had triggered without polling the sleep. This PR polls the sleep timer.
## Issue Addressed
We still ping peers that are considered in a disconnecting state
## Proposed Changes
Do not ping peers once we decide they are disconnecting
Upgrade logs about ignored rpc messages
## Additional Info
--
## Issue Addressed
N/A
## Proposed Changes
https://github.com/sigp/lighthouse/pull/3133 changed the rpc type limits to be fork aware i.e. if our current fork based on wall clock slot is Altair, then we apply only altair rpc type limits. This is a bug because phase0 blocks can still be sent over rpc and phase 0 block minimum size is smaller than altair block minimum size. So a phase0 block with `size < SIGNED_BEACON_BLOCK_ALTAIR_MIN` will return an `InvalidData` error as it doesn't pass the rpc types bound check.
This error can be seen when we try syncing pre-altair blocks with size smaller than `SIGNED_BEACON_BLOCK_ALTAIR_MIN`.
This PR fixes the issue by also accounting for forks earlier than current_fork in the rpc limits calculation in the `rpc_block_limits_by_fork` function. I decided to hardcode the limits in the function because that seemed simpler than calculating previous forks based on current fork and doing a min across forks. Adding a new fork variant is simple and can the limits can be easily checked in a review.
Adds unit tests and modifies the syncing simulator to check the syncing from across fork boundaries.
The syncing simulator's block 1 would always be of phase 0 minimum size (404 bytes) which is smaller than altair min block size (since block 1 contains no attestations).
## Issue Addressed
N/A
## Proposed Changes
Fix the upper bound for blocks by root responses to be equal to the max merge block size instead of altair.
Further make the rpc response limits fork aware.
## Proposed Changes
Add a `lighthouse db` command with three initial subcommands:
- `lighthouse db version`: print the database schema version.
- `lighthouse db migrate --to N`: manually upgrade (or downgrade!) the database to a different version.
- `lighthouse db inspect --column C`: log the key and size in bytes of every value in a given `DBColumn`.
This PR lays the groundwork for other changes, namely:
- Mark's fast-deposit sync (https://github.com/sigp/lighthouse/pull/2915), for which I think we should implement a database downgrade (from v9 to v8).
- My `tree-states` work, which already implements a downgrade (v10 to v8).
- Standalone purge commands like `lighthouse db purge-dht` per https://github.com/sigp/lighthouse/issues/2824.
## Additional Info
I updated the `strum` crate to 0.24.0, which necessitated some changes in the network code to remove calls to deprecated methods.
Thanks to @winksaville for the motivation, and implementation work that I used as a source of inspiration (https://github.com/sigp/lighthouse/pull/2685).
## Issue Addressed
#3006
## Proposed Changes
This PR changes the default behaviour of lighthouse to ignore discovered IPs that are not globally routable. It adds a CLI flag, --enable-local-discovery to permit the non-global IPs in discovery.
NOTE: We should take care in merging this as I will break current set-ups that rely on local IP discovery. I made this the non-default behaviour because we dont really want to be wasting resources attempting to connect to non-routable addresses and we dont want to propagate these to others (on the chance we can connect to one of these local nodes), improving discoveries efficiency.
## Proposed Changes
Lots of lint updates related to `flat_map`, `unwrap_or_else` and string patterns. I did a little more creative refactoring in the op pool, but otherwise followed Clippy's suggestions.
## Additional Info
We need this PR to unblock CI.
## Issue Addressed
I noticed in some logs some excess and unecessary discovery queries. What was happening was we were pruning our peers down to our outbound target and having some disconnect. When we are below this threshold we try to find more peers (even if we are at our peer limit). The request becomes futile because we have no more peer slots.
This PR corrects this issue and advances the pruning mechanism to favour subnet peers.
An overview the new logic added is:
- We prune peers down to a target outbound peer count which is higher than the minimum outbound peer count.
- We only search for more peers if there is room to do so, and we are below the minimum outbound peer count not the target. So this gives us some buffer for peers to disconnect. The buffer is currently 10%
The modified pruning logic is documented in the code but for reference it should do the following:
- Prune peers with bad scores first
- If we need to prune more peers, then prune peers that are subscribed to a long-lived subnet
- If we still need to prune peers, the prune peers that we have a higher density of on any given subnet which should drive for uniform peers across all subnets.
This will need a bit of testing as it modifies some significant peer management behaviours in lighthouse.
## Issue Addressed
NA
## Proposed Changes
This PR extends #3018 to address my review comments there and add automated integration tests with Geth (and other implementations, in the future).
I've also de-duplicated the "unused port" logic by creating an `common/unused_port` crate.
## Additional Info
I'm not sure if we want to merge this PR, or update #3018 and merge that. I don't mind, I'm primarily opening this PR to make sure CI works.
Co-authored-by: Mark Mackey <mark@sigmaprime.io>
## Issue Addressed
Lighthouse gossiping late messages
## Proposed Changes
Point LH to our fork using tokio interval, which 1) works as expected 2) is more performant than the previous version that actually worked as expected
Upgrade libp2p
## Additional Info
https://github.com/libp2p/rust-libp2p/issues/2497
## Issue Addressed
On a network with few nodes, it is possible that the same node can be found from a subnet discovery and a normal peer discovery at the same time.
The network behaviour loads these peers into events and processes them when it has the chance. It can happen that the same peer can enter the event queue more than once and then attempt to be dialed twice.
This PR shifts the registration of nodes in the peerdb as being dialed before they enter the NetworkBehaviour queue, preventing multiple attempts of the same peer being entered into the queue and avoiding the race condition.
## Issue Addressed
#2947
## Proposed Changes
Store messages that fail to be published due to insufficient peers for retry later. Messages expire after half an epoch and are retried if gossipsub informs us that an useful peer has connected. Currently running in Atlanta
## Additional Info
If on retry sending the messages fails they will not be tried again
The gossipsub history was increased to a good portion of a slot from 2.1 seconds in the last release.
Although it shouldn't cause too much issue, it could be related to recieving later messages than usual and interacting with our scoring system penalizing peers. For consistency, this PR reduces the time we gossip messages back to the same values of the previous release.
It also adjusts the gossipsub heartbeat time for testing purposes with a developer flag but this should not effect end users.
In the latest release we decreased the target number of subnet peers.
It appears this could be causing issues in some cases and so reverting it back to the previous number it wise. A larger PR that follows this will address some other related discovery issues and peer management around subnet peer discovery.
## Issue Addressed
We emit a warning to verify that all peer connection state information is consistent. A warning is given under one edge case;
We try to dial a peer with peer-id X and multiaddr Y. The peer responds to multiaddr Y with a different peer-id, Z. The dialing to the peer fails, but libp2p injects the failed attempt as peer-id Z.
In this instance, our PeerDB tries to add a new peer in the disconnected state under a previously unknown peer-id. This is harmless and so this PR permits this behaviour without logging a warning.
## Issue Addressed
The PeerDB was getting out of sync with the number of disconnected peers compared to the actual count. As this value determines how many we store in our cache, over time the cache was depleting and we were removing peers immediately resulting in errors that manifest as unknown peers for some operations.
The error occurs when dialing a peer fails, we were not correctly updating the peerdb counter because the increment to the counter was placed in the wrong order and was therefore not incrementing the count.
This PR corrects this.
There is a pretty significant tradeoff between bandwidth and speed of gossipsub messages.
We can reduce our bandwidth usage considerably at the cost of minimally delaying gossipsub messages. The impact of delaying messages has not been analyzed thoroughly yet, however this PR in conjunction with some gossipsub updates show considerable bandwidth reduction.
This PR allows the user to set a CLI value (`network-load`) which is an integer in the range of 1 of 5 depending on their bandwidth appetite. 1 represents the least bandwidth but slowest message recieving and 5 represents the most bandwidth and fastest received message time.
For low-bandwidth users it is likely to be more efficient to use a lower value. The default is set to 3, which currently represents a reduced bandwidth usage compared to previous version of this PR. The previous lighthouse versions are equivalent to setting the `network-load` CLI to 4.
This PR is awaiting a few gossipsub updates before we can get it into lighthouse.
## Issue Addressed
#2841
## Proposed Changes
Not counting dialing peers while deciding if we have reached the target peers in case of outbound peers.
## Additional Info
Checked this running in nodes and bandwidth looks normal, peer count looks normal too
Fix max packet sizes
Fix max_payload_size function
Add merge block test
Fix max size calculation; fix up test
Clear comments
Add a payload_size_function
Use safe arith for payload calculation
Return an error if block too big in block production
Separate test to check if block is over limit
Added Execution Payload from Rayonism Fork
Updated new Containers to match Merge Spec
Updated BeaconBlockBody for Merge Spec
Completed updating BeaconState and BeaconBlockBody
Modified ExecutionPayload<T> to use Transaction<T>
Mostly Finished Changes for beacon-chain.md
Added some things for fork-choice.md
Update to match new fork-choice.md/fork.md changes
ran cargo fmt
Added Missing Pieces in eth2_libp2p for Merge
fix ef test
Various Changes to Conform Closer to Merge Spec
## Issue Addressed
Users are experiencing `Status'd peer not found` errors
## Proposed Changes
Although I cannot reproduce this error, this is only one connection state change that is not addressed in the peer manager (that I could see). The error occurs because the number of disconnected peers in the peerdb becomes out of sync with the actual number of disconnected peers. From what I can tell almost all possible connection state changes are handled, except for the case when a disconnected peer changes to be disconnecting. This can potentially happen at the peer connection limit, where a previously connected peer switches to disconnecting.
This PR decrements the disconnected counter when this event occurs and from what I can tell, covers all possible disconnection state changes in the peer manager.
## Issue Addressed
N/A
## Proposed Changes
1. Don't disconnect peer from dht on connection limit errors
2. Bump up `PRIORITY_PEER_EXCESS` to allow for dialing upto 60 peers by default.
Co-authored-by: Diva M <divma@protonmail.com>
I had this change but it seems to have been lost in chaos of network upgrades.
The swarm dialing event seems to miss some cases where we dial via the behaviour. This causes an error to be logged as the peer manager doesn't know about some dialing events.
This shifts the logic to the behaviour to inform the peer manager.
