dummyHook's fields were concurrently written by nodes and read by
the test. The simplest solution is to protect all fields with a mutex.
Enable: TestMultiplePeersDropSelf, TestMultiplePeersDropOther as they
seemingly accidentally stayed disabled during a refactor/rewrite
since 1836366ac1.
resolvesethersphere/go-ethereum#1286
* p2p/discover: remove unused function
* p2p/enode: use localItemKey for local sequence number
I added localItemKey for this purpose in #18963, but then
forgot to actually use it. This changes the database layout
yet again and requires bumping the version number.
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.
I added localItemKey for this purpose in #18963, but then
forgot to actually use it. This changes the database layout
yet again and requires bumping the version number.
* swarm/network: DRY out repeated giga comment
I not necessarily agree with the way we wait for event propagation.
But I truly disagree with having duplicated giga comments.
* p2p/simulations: encapsulate Node.Up field so we avoid data races
The Node.Up field was accessed concurrently without "proper" locking.
There was a lock on Network and that was used sometimes to access
the field. Other times the locking was missed and we had
a data race.
For example: https://github.com/ethereum/go-ethereum/pull/18464
The case above was solved, but there were still intermittent/hard to
reproduce races. So let's solve the issue permanently.
resolves: ethersphere/go-ethereum#1146
* p2p/simulations: fix unmarshal of simulations.Node
Making Node.Up field private in 13292ee897e345045fbfab3bda23a77589a271c1
broke TestHTTPNetwork and TestHTTPSnapshot. Because the default
UnmarshalJSON does not handle unexported fields.
Important: The fix is partial and not proper to my taste. But I cut
scope as I think the fix may require a change to the current
serialization format. New ticket:
https://github.com/ethersphere/go-ethereum/issues/1177
* p2p/simulations: Add a sanity test case for Node.Config UnmarshalJSON
* p2p/simulations: revert back to defer Unlock() pattern for Network
It's a good patten to call `defer Unlock()` right after `Lock()` so
(new) error cases won't miss to unlock. Let's get back to that pattern.
The patten was abandoned in 85a79b3ad3,
while fixing a data race. That data race does not exist anymore,
since the Node.Up field got hidden behind its own lock.
* p2p/simulations: consistent naming for test providers Node.UnmarshalJSON
* p2p/simulations: remove JSON annotation from private fields of Node
As unexported fields are not serialized.
* p2p/simulations: fix deadlock in Network.GetRandomDownNode()
Problem: GetRandomDownNode() locks -> getDownNodeIDs() ->
GetNodes() tries to lock -> deadlock
On Network type, unexported functions must assume that `net.lock`
is already acquired and should not call exported functions which
might try to lock again.
* p2p/simulations: ensure method conformity for Network
Connect* methods were moved to p2p/simulations.Network from
swarm/network/simulation. However these new methods did not follow
the pattern of Network methods, i.e., all exported method locks
the whole Network either for read or write.
* p2p/simulations: fix deadlock during network shutdown
`TestDiscoveryPersistenceSimulationSimAdapter` often got into deadlock.
The execution was stuck on two locks, i.e, `Kademlia.lock` and
`p2p/simulations.Network.lock`. Usually the test got stuck once in each
20 executions with high confidence.
`Kademlia` was stuck in `Kademlia.EachAddr()` and `Network` in
`Network.Stop()`.
Solution: in `Network.Stop()` `net.lock` must be released before
calling `node.Stop()` as stopping a node (somehow - I did not find
the exact code path) causes `Network.InitConn()` to be called from
`Kademlia.SuggestPeer()` and that blocks on `net.lock`.
Related ticket: https://github.com/ethersphere/go-ethereum/issues/1223
* swarm/state: simplify if statement in DBStore.Put()
* p2p/simulations: remove faulty godoc from private function
The comment started with the wrong method name.
The method is simple and self explanatory. Also, it's private.
=> Let's just remove the comment.
* node: close AccountsManager in new Close method
* p2p/simulations, p2p/simulations/adapters: handle node close on shutdown
* node: move node ephemeralKeystore cleanup to stop method
* node: call Stop in Node.Close method
* cmd/geth: close node.Node created with makeFullNode in cli commands
* node: close Node instances in tests
* cmd/geth, node: minor code style fixes
* cmd, console, miner, mobile: proper node Close() termination
This change clears up confusion around the two ways in which nodes
can be added to the table.
When a neighbors packet is received as a reply to findnode, the nodes
contained in the reply are added as 'seen' entries if sufficient space
is available.
When a ping is received and the endpoint verification has taken place,
the remote node is added as a 'verified' entry or moved to the front of
the bucket if present. This also updates the node's IP address and port
if they have changed.
This change resolves multiple issues around handling of endpoint proofs.
The proof is now done separately for each IP and completing the proof
requires a matching ping hash.
Also remove waitping because it's equivalent to sleep. waitping was
slightly more efficient, but that may cause issues with findnode if
packets are reordered and the remote end sees findnode before pong.
Logging of received packets was hitherto done after handling the packet,
which meant that sent replies were logged before the packet that
generated them. This change splits up packet handling into 'preverify'
and 'handle'. The error from 'preverify' is logged, but 'handle' happens
after the message is logged. This fixes the order. Packet logs now
contain the node ID.
* p2p/simulation: WIP minimal snapshot test
* p2p/simulation: Add snapshot create, load and verify to snapshot test
* build: add test tag for tests
* p2p/simulations, build: Revert travis change, build test sym always
* p2p/simulations: Add comments, timeout check on additional events
* p2p/simulation: Add benchmark template for minimal peer protocol init
* p2p/simulations: Remove unused code
* p2p/simulation: Correct timer reset
* p2p/simulations: Put snapshot check events in buffer and call blocking
* p2p/simulations: TestSnapshot fail if Load function returns early
* p2p/simulations: TestSnapshot wait for all connections before returning
* p2p/simulation: Revert to before wait for snap load (5e75594)
* p2p/simulations: add "conns after load" subtest to TestSnapshot
and nudge
* swarm/network: Hive - do not notify peer if discovery is disabled
* p2p/simulations: validate all connections on loading a snapshot
* p2p/simulations: track all connections in on snapshot loading
* p2p/simulations: add snapshotLoadTimeout variable
* p2p/simulations: ignore control events in snapshot load
* p2p/simulations: simplify event loop synchronization
* p2p/simulations: return already connected error from Load function
* p2p/simulations: log warning on snapshot loading disconnection
This change extends the peer metrics collection:
- traces the life-cycle of the peers
- meters the peer traffic separately for every peer
- creates event feed for the peer events
- emits the peer events
This PR adds enode.LocalNode and integrates it into the p2p
subsystem. This new object is the keeper of the local node
record. For now, a new version of the record is produced every
time the client restarts. We'll make it smarter to avoid that in
the future.
There are a couple of other changes in this commit: discovery now
waits for all of its goroutines at shutdown and the p2p server
now closes the node database after discovery has shut down. This
fixes a leveldb crash in tests. p2p server startup is faster
because it doesn't need to wait for the external IP query
anymore.
This fixes a rare deadlock with the inproc adapter:
- A node is stopped, which acquires Network.lock.
- The protocol code being simulated (swarm/network in my case)
waits for its goroutines to shut down.
- One of those goroutines calls into the simulation to add a peer,
which waits for Network.lock.
The fix for the deadlock is really simple, just release the lock
before stopping the simulation node.
Other changes in this PR clean up the exec adapter so it reports
node startup errors better and remove the docker adapter because
it just adds overhead.
In the exec adapter, node information is now posted to a one-shot
server. This avoids log parsing and allows reporting startup
errors to the simulation host.
A small change in package node was needed because simulation
nodes use port zero. Node.{HTTP,WS}Endpoint now return the live
endpoints after startup by checking the TCP listener.
