We can reduce the size of test fixtures (which will improve test
reliability) without impacting these tests' primary role (which is
correctness.)
Also reducing these test logging will make the tests easier to read,
which whill be a good quality of life improvement for devs.
I see this panic in tests occasionally, and I don't think there's any
need to close this channel:
- it's only sent to in one place which has a select case with a
default clause, so there's no chance of deadlocks.
- the only place we recieve from it thas a timeout.
This contains two major changes:
- Remove the legacy test logging method, and just explicitly call the
noop logger. This is just to make the test logging behavior more
coherent and clear.
- Move the logging in the light package from the testing.T logger to
the noop logger. It's really the case that we very rarely need/want
to consider test logs unless we're doing reproductions and running a
narrow set of tests.
In most cases, I (for one) prefer to run in verbose mode so I can
watch progress of tests, but I basically never need to consider
logs. If I do want to see logs, then I can edit in the testing.T
logger locally (which is what you have to do today, anyway.)
This is failing intermittently, but it's a really simple test, and I
suspect that we're just running into thread scheduling issues on CI
nodes. I don't think making the test smaller reduces the utility of
this test.
closes: #7756
# What does this pull request change?
This pull request adds a new runbook for operators enountering errors related to the new Proposer-Based Timestamps algorithm. The goal of this runbook is to give operators a set of clear steps that they can follow if they are having issues producing blocks because of clock synchronization problems.
This pull request also renames the `*PrevoteDelay` metrics to drop the term `MessageDelay`. These metrics provide a combined view of `message_delay` + `synchrony` so the name may be confusing.
# Questions to reviewers
* Are there ways to make the set of steps clearer or are there any pieces that seem confusing?
This is minor, but I was trying to write a test and realized that the
application reference in the harness isn't actually used, which is
quite confusing.
The events switch code is largely vestigal and is responsible for
wiring between the consensus state machine and the consensus
reactor. While there might have been a need, historicallly to managed
these subscriptions at runtime, it's nolonger used: subscriptions are
registered during startup, and then the switch shuts down at at the
end.
Eventually the EventSwitch should be replaced by a much smaller
implementation of an eventloop in the consensus state machine, but
cutting down on the scope of the event switch will help clarify the
requirements from the consensus side.
This change implements the logic for the PrepareProposal ABCI++ method call. The main logic for creating and issuing the PrepareProposal request lives in execution.go and is tested in a set of new tests in execution_test.go. This change also updates the mempool mock to use a mockery generated version and removes much of the plumbing for the no longer used ABCIResponses.
* libs/log: remove Must constructor
* Update test/e2e/node/main.go
Co-authored-by: M. J. Fromberger <michael.j.fromberger@gmail.com>
* use stdlog
Co-authored-by: M. J. Fromberger <michael.j.fromberger@gmail.com>
Updates #8077. The panic handler for consensus currently attempts to effect a
clean shutdown, but this can leave a failed node running in an unknown state
for an arbitrary amount of time after the failure.
Since a panic at this point means consensus is already irrecoverably broken, we
should not allow the node to continue executing. After making a best effort to
shut down the writeahead log, re-panic to ensure the node will terminate before
any further state transitions are processed.
Even with this change, it is possible some transitions may occur while the
cleanup is happening. It might be preferable to abort unconditionally without
any attempt at cleanup.
Related changes:
- Clean up the creation of WAL directories.
- Filter WAL close errors at rethrow.
The message handling in this reactor is all under control of the reactor
itself, and does not call out to callbacks or other externally-supplied code.
It doesn't need to check for panics.
- Remove an irrelevant channel ID check.
- Remove an unnecessary panic recovery wrapper.
The PEX reactor has a simple feedback control mechanism to decide how often to
poll peers for peer address updates. The idea is to poll more frequently when
knowledge of the network is less, and decrease frequency as knowledge grows.
This change solves two problems:
1. It is possible in some cases we may poll a peer "too often" and get dropped
by that peer for spamming.
2. The first successful peer update with any content resets the polling timer
to a very long time (10m), meaning if we are unlucky in getting an
incomplete reply while the network is small, we may not try again for a very
long time. This may contribute to difficulties bootstrapping sync.
The main change here is to only update the interval when new information is
added to the system, and not (as before) whenever a request is sent out to a
peer. The rate computation is essentially the same as before, although the code
has been a bit simplified, and I consolidated some of the error handling so
that we don't have to check in multiple places for the same conditions.
Related changes:
- Improve error diagnostics for too-soon and overflow conditions.
- Clean up state handling in the poll interval computation.
- Pin the minimum interval avert a chance of PEX spamming a peer.
This change set implements the most recent version of `FinalizeBlock`.
# What does this change actually contain?
* This change set is rather large but fear not! The majority of the files touched and changes are renaming `ResponseDeliverTx` to `ExecTxResult`. This should be a pretty inoffensive change since they're effectively the same type but with a different name.
* The `execBlockOnProxyApp` was totally removed since it served as just a wrapper around the logic that is now mostly encapsulated within `FinalizeBlock`
* The `updateState` helper function has been made a public method on `State`. It was being exposed as a shim through the testing infrastructure, so this seemed innocuous.
* Tests already existed to ensure that the application received the `ByzantineValidators` and the `ValidatorUpdates`, but one was fixed up to ensure that `LastCommitInfo` was being sent across.
* Tests were removed from the `psql` indexer that seemed to search for an event in the indexer that was not being created.
# Questions for reviewers
* We store this [ABCIResponses](5721a13ab1/proto/tendermint/state/types.pb.go (L37)) type in the data base as the block results. This type has changed since v0.35 to contain the `FinalizeBlock` response. I'm wondering if we need to do any shimming to keep the old data retrieveable?
* Similarly, this change is exposed via the RPC through [ResultBlockResults](5721a13ab1/rpc/coretypes/responses.go (L69)) changing. Should we somehow shim or notify for this change?
closes: #7658
Since the goal of reading events at the head of the event log is to satisfy a
subscription style interface, there is no point in allowing head polling with
no wait interval. The pagination case already bypasses long polling, so the
extra option is unneessary.
Set a minimum default long-polling interval for the head case.
Add a test for minimum delay.
This is mostly an extremely small change where I double a somewhat
arbitrarly set timeout from 1m to 2m for an entire test. When I put
these timeouts in the test, they were arbitrary based on my local
performance (which is quite fact,) and I expected that they'd need to
be tweaked in the future.
A big chunk of this PR is reworking a collection of helper functions
that produce somewhat intractable messages when a test fails, so that
the error messages take up less vertical space, hopefully without
losing any debugability.
We're waiting between trying witnesses (which shouldn't be neccessary
because the witnesses shouldn't depend on each other,) and also
between *attempts*, and really the outer sleep should be enough.
This is a little coarse, but the idea is that we'll send information
about the channels a peer has upon the peer-up event that we send to
reactors that we can then use to reject peers (if neeeded) from reactors.
This solves the problem where statesync would hang in test networks
(and presumably real) where we would attempt to statesync from seed
nodes, thereby hanging silently forever.
This change implements the spec for `ProcessProposal`. It first calls the Tendermint block validation logic to check that all of the proposed block fields are well formed and do not violate any of the rules for Tendermint to consider the block valid and then passes the validated block the `ProcessProposal`.
This change also adds additional fixtures to test the change. It adds the `baseMock` types that holds a mock as well as a reference to `BaseApplication`. If the function was not setup by the test on the contained mock Application, the type delegates to the `BaseApplication` and returns what `BaseApplication` returns.
The change also switches the `makeState` helper to take an arg struct so that an ABCI application can be plumbed through when needed.
closes: #7656
Sam Kleinman
William Banfield
JayT106
M. J. Fromberger