This continues the push of plumbing contexts through tendermint. I
attempted to find all goroutines in the production code (non-test) and
made sure that these threads would exit when their contexts were
canceled, and I believe this PR does that.
This is, perhaps, the trival final piece of #7075 that I've been
working on.
There's more work to be done:
- push more of the setup into the pacakges themselves
- move channel-based sending/filtering out of the
- simplify the buffering throuhgout the p2p stack.
Addresses one of the concerns with #7041.
Provides a mechanism (via the RPC interface) to delete a single transaction, described by its hash, from the mempool. The method returns an error if the transaction cannot be found. Once the transaction is removed it remains in the cache and cannot be resubmitted until the cache is cleared or it expires from the cache.
The code in the Tendermint repository makes heavy use of import aliasing.
This is made necessary by our extensive reuse of common base package names, and
by repetition of similar names across different subdirectories.
Unfortunately we have not been very consistent about which packages we alias in
various circumstances, and the aliases we use vary. In the spirit of the advice
in the style guide and https://github.com/golang/go/wiki/CodeReviewComments#imports,
his change makes an effort to clean up and normalize import aliasing.
This change makes no API or behavioral changes. It is a pure cleanup intended
o help make the code more readable to developers (including myself) trying to
understand what is being imported where.
Only unexported names have been modified, and the changes were generated and
applied mechanically with gofmt -r and comby, respecting the lexical and
syntactic rules of Go. Even so, I did not fix every inconsistency. Where the
changes would be too disruptive, I left it alone.
The principles I followed in this cleanup are:
- Remove aliases that restate the package name.
- Remove aliases where the base package name is unambiguous.
- Move overly-terse abbreviations from the import to the usage site.
- Fix lexical issues (remove underscores, remove capitalization).
- Fix import groupings to more closely match the style guide.
- Group blank (side-effecting) imports and ensure they are commented.
- Add aliases to multiple imports with the same base package name.
This changes adds a failing test for issue #6660. It achieves this by adding a transaction, starting the `broadcastTxRoutine` in a goroutine and then adding another transaction to the mempool. The `broadcastTxRoutine` can receive the second inserted transaction before `insertTx` returns. In that case, `broadcastTxRoutine` will derefence a nil pointer when referencing the `gossipEl` and panic.
## Description
Internalize some libs. This reduces the amount ot public API tendermint is supporting. The moved libraries are mainly ones that are used within Tendermint-core.
This cleans up the `Router` code and adds a bunch of tests. These sorts of systems are a real pain to test, since they have a bunch of asynchronous goroutines living their own lives, so the test coverage is decent but not fantastic. Luckily we've been able to move all of the complex peer management and transport logic outside of the router, as synchronous components that are much easier to test, so the core router logic is fairly small and simple.
This also provides some initial test tooling in `p2p/p2ptest` that automatically sets up in-memory networks and channels for use in integration tests. It also includes channel-oriented test asserters in `p2p/p2ptest/require.go`, but these have primarily been written for router testing and should probably be adapted or extended for reactor testing.
This improves the `peerStore` prototype by e.g.:
* Using a database with Protobuf for persistence, but also keeping full peer set in memory for performance.
* Simplifying the API, by taking/returning struct copies for safety, and removing errors for in-memory operations.
* Caching the ranked peer set, as a temporary solution until a better data structure is implemented.
* Adding `PeerManagerOptions.MaxPeers` and pruning the peer store (based on rank) when it's full.
* Rewriting `PeerAddress` to be independent of `url.URL`, normalizing it and tightening semantics.
@p4u from vocdoni.io reported that the mempool might behave incorrectly under a
high load. The consequences can range from pauses between blocks to the peers
disconnecting from this node.
My current theory is that the flowrate lib we're using to control flow
(multiplex over a single TCP connection) was not designed w/ large blobs
(1MB batch of txs) in mind.
I've tried decreasing the Mempool reactor priority, but that did not
have any visible effect. What actually worked is adding a time.Sleep
into mempool.Reactor#broadcastTxRoutine after an each successful send ==
manual control flow of sort.
As a temporary remedy (until the mempool package
is refactored), the max-batch-bytes was disabled. Transactions will be sent
one by one without batching
Closes#5796
After a reactor has failed to parse an incoming message, it shouldn't output the "bad" data into the logs, as that data is unfiltered and could have anything in it. (We also don't think this information is helpful to have in the logs anyways.)
## Description
This PR wraps the stdlib sync.(RW)Mutex & godeadlock.(RW)Mutex. This enables using go-deadlock via a build flag instead of using sed to replace sync with godeadlock in all files
Closes: #3242
## Description
To provide the ability to add more message types without needing to cause a breaking change the mempool message was migrated to a oneof.
Closes: #XXX
* proto: move mempool to proto
- changes according to moving the mempool reactor to proto
Signed-off-by: Marko Baricevic <marbar3778@yahoo.com>
Closes: #2883
to prevent malicious nodes from sending us large messages (~21MB, which
is the default `RecvMessageCapacity`)
This allows us to remove unnecessary `maxMsgSize` check in `decodeMsg`. Since each channel has a msg capacity set to `maxMsgSize`, there's no need to check it again in `decodeMsg`.
Closes#1503
Previously, many reactors were initialized with the name "Reactor," which made it difficult to log which reactor was doing what. This changes those reactors' names to something more descriptive.
* Include sender when logging rejected txns
* Log as peerID to be consistent with other log messages
* Updated CHANGELOG_PENDING
* Handle nil source
* Updated PR link in CHANGELOG_PENDING
* Renamed TxInfo.SenderAddress and peerAddress til PeerFullID
* Renamed PeerFullID to PeerP2PID
* Forgot to rename a couple of references
## Description
Refs #2659
Breaking changes in the mempool package:
[mempool] #2659 Mempool now an interface
old Mempool renamed to CListMempool
NewMempool renamed to NewCListMempool
Option renamed to CListOption
MempoolReactor renamed to Reactor
NewMempoolReactor renamed to NewReactor
unexpose TxID method
TxInfo.PeerID renamed to SenderID
unexpose MempoolReactor.Mempool
Breaking changes in the state package:
[state] #2659 Mempool interface moved to mempool package
MockMempool moved to top-level mock package and renamed to Mempool
Non Breaking changes in the node package:
[node] #2659 Add Mempool method, which allows you to access mempool
## Commits
* move Mempool interface into mempool package
Refs #2659
Breaking changes in the mempool package:
- Mempool now an interface
- old Mempool renamed to CListMempool
Breaking changes to state package:
- MockMempool moved to mempool/mock package and renamed to Mempool
- Mempool interface moved to mempool package
* assert CListMempool impl Mempool
* gofmt code
* rename MempoolReactor to Reactor
- combine everything into one interface
- rename TxInfo.PeerID to TxInfo.SenderID
- unexpose MempoolReactor.Mempool
* move mempool mock into top-level mock package
* add a fixme
TxsFront should not be a part of the Mempool interface
because it leaks implementation details. Instead, we need to come up
with general interface for querying the mempool so the MempoolReactor
can fetch and broadcast txs to peers.
* change node#Mempool to return interface
* save commit = new reactor arch
* Revert "save commit = new reactor arch"
This reverts commit 1bfceacd9d.
* require CListMempool in mempool.Reactor
* add two changelog entries
* fixes after my own review
* quote interfaces, structs and functions
* fixes after Ismail's review
* make node's mempool an interface
* make InitWAL/CloseWAL methods a part of Mempool interface
* fix merge conflicts
* make node's mempool an interface
* mempool: add a safety check, write tests for mempoolIDs
and document 65536 limit in the mempool reactor spec
follow-up to https://github.com/tendermint/tendermint/pull/2778
* rename the test
* fixes after Ismail's review
Refs #3306, irisnet@fdbb676
I ran an irishub validator. After the validator node ran several days, I dump the whole goroutine stack. I found that there were hundreds of broadcastTxRoutine. However, the connected peer quantity was less than 30. So I belive that there must be broadcastTxRoutine leakage issue.
According to my analysis, I think the root cause of this issue locate in below code:
select {
case <-next.NextWaitChan():
// see the start of the for loop for nil check
next = next.Next()
case <-peer.Quit():
return
case <-memR.Quit():
return
}
As we know, if multiple paths are avaliable in the same time, then a random path will be selected. Suppose that next.NextWaitChan() and peer.Quit() are both avaliable, and next.NextWaitChan() is chosen.
// send memTx
msg := &TxMessage{Tx: memTx.tx}
success := peer.Send(MempoolChannel, cdc.MustMarshalBinaryBare(msg))
if !success {
time.Sleep(peerCatchupSleepIntervalMS * time.Millisecond)
continue
}
Then next will be non-empty and the peer send operation won't be success. As a result, this go routine will be track into infinite loop and won't be released.
My proposal is to check peer.Quit() and memR.Quit() in every loop no matter whether next is nil.
Closes#1798
This is done by making every mempool tx maintain a list of peers who its received the tx from. Instead of using the 20byte peer ID, it instead uses a local map from peerID to uint16 counter, so every peer adds 2 bytes. (Word aligned to probably make it 8 bytes)
This also required resetting the callback function on every CheckTx. This likely has performance ramifications for instruction caching. The actual setting operation isn't costly with the removal of defers in this PR.
* Make the mempool not gossip txs back to peers its received it from
* Fix adversarial memleak
* Don't break interface
* Update changelog
* Forgot to add a mtx
* forgot a mutex
* Update mempool/reactor.go
Co-Authored-By: ValarDragon <ValarDragon@users.noreply.github.com>
* Update mempool/mempool.go
Co-Authored-By: ValarDragon <ValarDragon@users.noreply.github.com>
* Use unknown peer ID
Co-Authored-By: ValarDragon <ValarDragon@users.noreply.github.com>
* fix compilation
* use next wait chan logic when skipping
* Minor fixes
* Add TxInfo
* Add reverse map
* Make activeID's auto-reserve 0
* 0 -> UnknownPeerID
Co-Authored-By: ValarDragon <ValarDragon@users.noreply.github.com>
* Switch to making the normal case set a callback on the reqres object
The recheck case is still done via the global callback, and stats
are also set via global callback
* fix merge conflict
* Addres comments
* Add cache tests
* add cache tests
* minor fixes
* update metrics in reqResCb and reformat code
* goimport -w mempool/reactor.go
* mempool: update memTx senders
I had to introduce txsMap for quick mempoolTx lookups.
* change senders type from []uint16 to sync.Map
Fixes DATA RACE:
```
Read at 0x00c0013fcd3a by goroutine 183:
github.com/tendermint/tendermint/mempool.(*MempoolReactor).broadcastTxRoutine()
/go/src/github.com/tendermint/tendermint/mempool/reactor.go:195 +0x3c7
Previous write at 0x00c0013fcd3a by D[2019-02-27|10:10:49.058] Read PacketMsg switch=3 peer=35bc1e3558c182927b31987eeff3feb3d58a0fc5@127.0.0.1
:46552 conn=MConn{pipe} packet="PacketMsg{30:2B06579D0A143EB78F3D3299DE8213A51D4E11FB05ACE4D6A14F T:1}"
goroutine 190:
github.com/tendermint/tendermint/mempool.(*Mempool).CheckTxWithInfo()
/go/src/github.com/tendermint/tendermint/mempool/mempool.go:387 +0xdc1
github.com/tendermint/tendermint/mempool.(*MempoolReactor).Receive()
/go/src/github.com/tendermint/tendermint/mempool/reactor.go:134 +0xb04
github.com/tendermint/tendermint/p2p.createMConnection.func1()
/go/src/github.com/tendermint/tendermint/p2p/peer.go:374 +0x25b
github.com/tendermint/tendermint/p2p/conn.(*MConnection).recvRoutine()
/go/src/github.com/tendermint/tendermint/p2p/conn/connection.go:599 +0xcce
Goroutine 183 (running) created at:
D[2019-02-27|10:10:49.058] Send switch=2 peer=1efafad5443abeea4b7a8155218e4369525d987e@127.0.0.1:46193 channel=48 conn=MConn{pipe} m
sgBytes=2B06579D0A146194480ADAE00C2836ED7125FEE65C1D9DD51049
github.com/tendermint/tendermint/mempool.(*MempoolReactor).AddPeer()
/go/src/github.com/tendermint/tendermint/mempool/reactor.go:105 +0x1b1
github.com/tendermint/tendermint/p2p.(*Switch).startInitPeer()
/go/src/github.com/tendermint/tendermint/p2p/switch.go:683 +0x13b
github.com/tendermint/tendermint/p2p.(*Switch).addPeer()
/go/src/github.com/tendermint/tendermint/p2p/switch.go:650 +0x585
github.com/tendermint/tendermint/p2p.(*Switch).addPeerWithConnection()
/go/src/github.com/tendermint/tendermint/p2p/test_util.go:145 +0x939
github.com/tendermint/tendermint/p2p.Connect2Switches.func2()
/go/src/github.com/tendermint/tendermint/p2p/test_util.go:109 +0x50
I[2019-02-27|10:10:49.058] Added good transaction validator=0 tx=43B4D1F0F03460BD262835C4AA560DB860CFBBE85BD02386D83DAC38C67B3AD7 res="&{CheckTx:gas_w
anted:1 }" height=0 total=375
Goroutine 190 (running) created at:
github.com/tendermint/tendermint/p2p/conn.(*MConnection).OnStart()
/go/src/github.com/tendermint/tendermint/p2p/conn/connection.go:210 +0x313
github.com/tendermint/tendermint/libs/common.(*BaseService).Start()
/go/src/github.com/tendermint/tendermint/libs/common/service.go:139 +0x4df
github.com/tendermint/tendermint/p2p.(*peer).OnStart()
/go/src/github.com/tendermint/tendermint/p2p/peer.go:179 +0x56
github.com/tendermint/tendermint/libs/common.(*BaseService).Start()
/go/src/github.com/tendermint/tendermint/libs/common/service.go:139 +0x4df
github.com/tendermint/tendermint/p2p.(*peer).Start()
<autogenerated>:1 +0x43
github.com/tendermint/tendermint/p2p.(*Switch).startInitPeer()
```
* explain the choice of a map DS for senders
* extract ids pool/mapper to a separate struct
* fix literal copies lock value from senders: sync.Map contains sync.Mutex
* use sync.Map#LoadOrStore instead of Load
* fixes after Ismail's review
* rename resCbNormal to resCbFirstTime
Sam Kleinman
M. J. Fromberger
William Banfield
Aleksandr Bezobchuk
Marko
Erik Grinaker
Anton Kaliaev
Tess Rinearson
Erik Grinaker
Jun Kimura
HaoyangLiu
Dev Ojha