This PR fixes multiple issues with the UDP connection pre-negotiation feature:
- the enable condition was wrong (it checked the existence of the DiscV5 struct where it wasn't initialized yet, disabling the feature even if discv5 was enabled)
- the server pool queried already connected nodes when the discovery iterators returned them again
- servers responded positively before they were synced and really willing to accept connections
Metrics are also added on the server side that count the positive and negative replies to served connection queries.
This PR fixes multiple issues with the UDP connection pre-negotiation feature:
- the enable condition was wrong (it checked the existence of the DiscV5 struct where it wasn't initialized yet, disabling the feature even if discv5 was enabled)
- the server pool queried already connected nodes when the discovery iterators returned them again
- servers responded positively before they were synced and really willing to accept connections
Metrics are also added on the server side that count the positive and negative replies to served connection queries.
This PR implements the first one of the "lespay" UDP queries which
is already useful in itself: the capacity query. The server pool is making
use of this query by doing a cheap UDP query to determine whether it is
worth starting the more expensive TCP connection process.
* les: move serverPool to les/vflux/client
* les: add metrics
* les: moved ValueTracker inside ServerPool
* les: protect against node registration before server pool is started
* les/vflux/client: fixed tests
* les: make peer registration safe
* les: refactored server handler
* tests/fuzzers/les: add fuzzer for les server handler
* tests, les: update les fuzzer
tests: update les fuzzer
tests/fuzzer/les: release resources
tests/fuzzer/les: pre-initialize all resources
* les: refactored server handler and fuzzer
Co-authored-by: rjl493456442 <garyrong0905@gmail.com>
This PR enables running the new discv5 protocol in both LES client
and server mode. In client mode it mixes discv5 and dnsdisc iterators
(if both are enabled) and filters incoming ENRs for "les" tag and fork ID.
The old p2p/discv5 package and all references to it are removed.
Co-authored-by: Felix Lange <fjl@twurst.com>
* les: allow tx unindexing in les/4 light server mode
* les: minor fixes
* les: more small fixes
* les: add meaningful constants for recentTxIndex handshake field
This PR adds an extra guarantee to NodeStateMachine: it ensures that all
immediate effects of a certain change are processed before any subsequent
effects of any of the immediate effects on the same node. In the original
version, if a cascaded change caused a subscription callback to be called
multiple times for the same node then these calls might have happened in a
wrong chronological order.
For example:
- a subscription to flag0 changes flag1 and flag2
- a subscription to flag1 changes flag3
- a subscription to flag1, flag2 and flag3 was called in the following order:
[flag1] -> [flag1, flag3]
[] -> [flag1]
[flag1, flag3] -> [flag1, flag2, flag3]
This happened because the tree of changes was traversed in a "depth-first
order". Now it is traversed in a "breadth-first order"; each node has a
FIFO queue for pending callbacks and each triggered subscription callback
is added to the end of the list. The already existing guarantees are
retained; no SetState or SetField returns until the callback queue of the
node is empty again. Just like before, it is the responsibility of the
state machine design to ensure that infinite state loops are not possible.
Multiple changes affecting the same node can still happen simultaneously;
in this case the changes can be interleaved in the FIFO of the node but the
correct order is still guaranteed.
A new unit test is also added to verify callback order in the above scenario.
This PR reimplements the light client server pool. It is also a first step
to move certain logic into a new lespay package. This package will contain
the implementation of the lespay token sale functions, the token buying and
selling logic and other components related to peer selection/prioritization
and service quality evaluation. Over the long term this package will be
reusable for incentivizing future protocols.
Since the LES peer logic is now based on enode.Iterator, it can now use
DNS-based fallback discovery to find servers.
This document describes the function of the new components:
https://gist.github.com/zsfelfoldi/3c7ace895234b7b345ab4f71dab102d4
This PR adds service value measurement statistics to the light client. It
also adds a private API that makes these statistics accessible. A follow-up
PR will add the new server pool which uses these statistics to select
servers with good performance.
This document describes the function of the new components:
https://gist.github.com/zsfelfoldi/3c7ace895234b7b345ab4f71dab102d4
Co-authored-by: rjl493456442 <garyrong0905@gmail.com>
Co-authored-by: rjl493456442 <garyrong0905@gmail.com>
This change
- implements concurrent LES request serving even for a single peer.
- replaces the request cost estimation method with a cost table based on
benchmarks which gives much more consistent results. Until now the
allowed number of light peers was just a guess which probably contributed
a lot to the fluctuating quality of available service. Everything related
to request cost is implemented in a single object, the 'cost tracker'. It
uses a fixed cost table with a global 'correction factor'. Benchmark code
is included and can be run at any time to adapt costs to low-level
implementation changes.
- reimplements flowcontrol.ClientManager in a cleaner and more efficient
way, with added capabilities: There is now control over bandwidth, which
allows using the flow control parameters for client prioritization.
Target utilization over 100 percent is now supported to model concurrent
request processing. Total serving bandwidth is reduced during block
processing to prevent database contention.
- implements an RPC API for the LES servers allowing server operators to
assign priority bandwidth to certain clients and change prioritized
status even while the client is connected. The new API is meant for
cases where server operators charge for LES using an off-protocol mechanism.
- adds a unit test for the new client manager.
- adds an end-to-end test using the network simulator that tests bandwidth
control functions through the new API.
This PR enables the indexers to work in light client mode by
downloading a part of these tries (the Merkle proofs of the last
values of the last known section) in order to be able to add new
values and recalculate subsequent hashes. It also adds CHT data to
NodeInfo.
This PR implements les.freeClientPool. It also adds a simulated clock
in common/mclock, which enables time-sensitive tests to run quickly
and still produce accurate results, and package common/prque which is
a generalised variant of prque that enables removing elements other
than the top one from the queue.
les.freeClientPool implements a client database that limits the
connection time of each client and manages accepting/rejecting
incoming connections and even kicking out some connected clients. The
pool calculates recent usage time for each known client (a value that
increases linearly when the client is connected and decreases
exponentially when not connected). Clients with lower recent usage are
preferred, unknown nodes have the highest priority. Already connected
nodes receive a small bias in their favor in order to avoid accepting
and instantly kicking out clients.
Note: the pool can use any string for client identification. Using
signature keys for that purpose would not make sense when being known
has a negative value for the client. Currently the LES protocol
manager uses IP addresses (without port address) to identify clients.
This PR fixes a retriever logic bug. When a peer had a soft timeout
and then a response arrived, it always assumed it was the same peer
even though it could have been a later requested one that did not time
out at all yet. In this case the logic went to an illegal state and
deadlocked, causing a goroutine leak.
Fixes#16243 and replaces #16359.
Thanks to @riceke for finding the bug in the logic.
This commit affects p2p/discv5 "topic discovery" by running it on
the same UDP port where the old discovery works. This is realized
by giving an "unhandled" packet channel to the old v4 discovery
packet handler where all invalid packets are sent. These packets
are then processed by v5. v5 packets are always invalid when
interpreted by v4 and vice versa. This is ensured by adding one
to the first byte of the packet hash in v5 packets.
DiscoveryV5Bootnodes is also changed to point to new bootnodes
that are implementing the changed packet format with modified
hash. Existing and new v5 bootnodes are both running on different
ports ATM.
This PR implements the new LES protocol version extensions:
* new and more efficient Merkle proofs reply format (when replying to
a multiple Merkle proofs request, we just send a single set of trie
nodes containing all necessary nodes)
* BBT (BloomBitsTrie) works similarly to the existing CHT and contains
the bloombits search data to speed up log searches
* GetTxStatusMsg returns the inclusion position or the
pending/queued/unknown state of a transaction referenced by hash
* an optional signature of new block data (number/hash/td) can be
included in AnnounceMsg to provide an option for "very light
clients" (mobile/embedded devices) to skip expensive Ethash check
and accept multiple signatures of somewhat trusted servers (still a
lot better than trusting a single server completely and retrieving
everything through RPC). The new client mode is not implemented in
this PR, just the protocol extension.
This commit does various code refactorings:
- generalizes and moves the request retrieval/timeout/resend logic out of LesOdr
(will be used by a subsequent PR)
- reworks the peer management logic so that all services can register with
peerSet to get notified about added/dropped peers (also gets rid of the ugly
getAllPeers callback in requestDistributor)
- moves peerSet, LesOdr, requestDistributor and retrieveManager initialization
out of ProtocolManager because I believe they do not really belong there and the
whole init process was ugly and ad-hoc
Felföldi Zsolt
