Do not dispatch a transaction received from a peer for
processing, if it has already been dispatched within the
past ten seconds.
Increase the number of transaction handlers that can be in
flight in the job queue and decrease the relative cost for
peers to share transaction and ledger data.
Additionally, make better use of resources by adjusting the
number of threads we initialize, by reverting commit
68b8ffdb63.
Both Tickets and SHAMapV2 have been around for a while and don't
look like they will be enabled on the network soon. So they are
removed from the supportedAmendments list. This prevents Env
from automatically testing with Tickets or SHAMapV2 enabled,
although testing with those features can still be explicitly
specified.
Drive-by cleanups:
o supportedAmendments() returns a const reference rather than
a fresh vector on each call.
o supportedAmendments() implementation moved from Amendments.cpp
to Feature.cpp. Amendments.cpp deleted.
o supportedAmendments() declared in Feature.h. All other
declarations deleted.
o preEnabledAmendments() removed, since it was empty and only
used in one place. It will be easy to re-add when it is needed.
o jtx::all_features_except() renamed to
jtx::supported_features_except(), which is more descriptive.
o jtx::all_amendments() renamed to jxt::supported_amendments()
o jtx::with_features() renamed to with_only_features()
o Env_test.cpp adjusted since featureTickets is no longer
automatically enabled for unit tests.
* If the transaction can't be queued, recover to the open ledger once,
and drop it on the next attempt.
* New result codes for transactions that can not queue.
* Add minimum queue size.
* Remove the obsolete and incorrect SF_RETRY flag.
* fix#2215
The two active users of DeadlineTimer, NetworkOPs and Application,
now use asio::steady_timers rather than DeadlineTimer.
DeadlineTimer is removed since it is no longer used.
To assure that all in-flight closures on timers are done before
Stoppables call stopped(), the JobCounter is made more generic.
It's now a ClosureCounter. The ClosureCounter is currently used
to count closures in flight for the JobQueue, NetworkOPs, and the
Application.
If the JobQueue is used during shutdown then those Jobs may access
Stoppables after they have already stopped. This violates the
preconditions of Stoppables and may lead to undefined behavior.
The solution taken here is to reference count all Jobs in the
JobQueue. At stop time all Jobs already in the JobQueue are
allowed to run to completion, but no further Jobs are allowed
into the JobQueue.
If a Job is rejected from the JobQueue (because we are stopping),
then JobQueue::addJob() returns false, so the caller can make any
necessary adjustments.
Introduces a generic Validations class for storing and querying current and
recent validations. Aditionally migrates the validation related timing
constants from LedgerTiming to the new Validations code.
Introduces RCLValidations as the version of Validations adapted for use in the
RCL. This adds support for flushing/writing validations to the sqlite log and
also manages concurrent access to the Validations data.
RCLValidations::flush() no longer uses the JobQueue for its database
write at shutdown. It performs the write directly without
changing threads.
The job queue can automatically tune the number of threads that
it creates based on the number of processors or processor cores
that are available.
The existing tuning was very conservative, limiting the maximum
number of threads to only 6.
Adjust the new algorithm to allow a larger number of threads and
allow server administrators to override the value in the config
file.
The ledger_request RPC call, under some conditions, did not
actually check that the entire ledger was present in the
database, making it unsuitable for use in cases where the
database was believed to be incorrect or incomplete.
With this change, the full ledger will be checked for
integrity unless it has already recently been checked
(according to the InboundLedgers cache).
This is a substantial refactor of the consensus code and also introduces
a basic consensus simulation and testing framework. The new generic/templated
version is in src/ripple/consensus and documents the current type requirements.
The version adapted for the RCL is in src/ripple/app/consensus. The testing
framework is in src/test/csf.
Minor behavioral changes/fixes include:
* Adjust close time offset even when not validating.
* Remove spurious proposing_ = false call at end of handleLCL.
* Remove unused functionality provided by checkLastValidation.
* Separate open and converge time
* Don't send a bow out if we're not proposing
* Prevent consensus stopping if NetworkOPs switches to disconnect mode while
consensus accepts a ledger
* Prevent a corner case in which Consensus::gotTxSet or Consensus::peerProposal
has the potential to update internal state while an dispatched accept job is
running.
* Distinguish external and internal calls to startNewRound. Only external
calls can reset the proposing_ state of consensus
The DatabaseImp has threads that asynchronously call JobQueue to
perform database reads. Formerly these threads had the same
lifespan as Database, which was until the end-of-life of
ApplicationImp. During shutdown these threads could call JobQueue
after JobQueue had already stopped. Or, even worse, occasionally
call JobQueue after JobQueue's destructor had run.
To avoid these shutdown conditions, Database is made a Stoppable,
with JobQueue as its parent. When Database stops, it shuts down
its asynchronous read threads. This prevents Database from
accessing JobQueue after JobQueue has stopped, but allows
Database to perform stores for the remainder of shutdown.
During development it was noted that the Database::close()
method was never called. So that method is removed from Database
and all derived classes.
Stoppable is also adjusted so it can be constructed using either
a char const* or a std::string.
For those files touched for other reasons, unneeded #includes
are removed.
Modify doStart Application method to specify whether or not to start the
DeadlineTimers. Specify inactive timers for jtx::Env Applications and
active timers for standard Applications.
Allow manifest revoking validator keys to be stored in a separate
[validator_key_revocation] config field, so the validator can run
again with new keys and token.
Validator lists from configured remote sites are fetched at a regular
interval. Fetched lists are expected to be in JSON format and contain the
following fields:
* "manifest": Base64-encoded serialization of a manifest containing the
validator publisher's master and signing public keys.
* "blob": Base64-encoded JSON string containing a "sequence",
"expiration" and "validators" field. "expiration" contains the Ripple
timestamp (seconds since January 1st, 2000 (00:00 UTC)) for when the
list expires. "validators" contains an array of objects with a
"validation_public_key" field.
* "signature": Hex-encoded signature of the blob using the publisher's
signing key.
* "version": 1
* "refreshInterval" (optional)
Instead of specifying a static list of trusted validators in the config
or validators file, the configuration can now include trusted validator
list publisher keys.
The trusted validator list and quorum are now reset each consensus
round using the latest validator lists and the list of recent
validations seen. The minimum validation quorum is now only
configurable via the command line.
When started with "--start", put all known, non-vetoed
amendments in the genesis ledger. This avoids the need
to wait 256 ledgers before amendments are enabled when
testing with a fresh ledger.
* Remove extraneous passing of transaction set hashes
* Remove recentPositions_. InboundTXs does the job now
* Move responsibility for sending "have TX set" out of consensus
cmake support in rippled. Currently supports:
* unity/nounity debug/release
* running protobuf
* sanitizer builds
* optional release build with assert turned on
* `target` variable to easily set compiler/debug/unity
(i.e. -Dtarget=gcc.debug.nounity)
* gcc/clang/visual studio/xcode
* linux/mac/win
* gcc 4 ABI, when needed
* ninja builds
* check openssl for acceptably recent release
* static builds
TBD:
* jemalloc support
* count
Notes:
* Use the -G"Visual Studio 14 2015 Win64" generator on Windows. Without
this a 32-bit project will be created. There is no way to set the
generator or force a 64-bit build in CMakeLists.txt (setting
CMAKE_GENERATOR_PLATFORM won't work). The best solution may be to
wrap cmake with a script.
* It is not possible to generate a visual studio project on linux or
mac. The visual studio generator is only available on windows.
* The visual studio project can be _either_ unity or
non-unity (selected at generation time). It does not appear possible
to disable compilation based on configuration.
* Language is _much_ worse than python, poor documentation and "quirky"
language support (for example, generator expressions can only be used
in limited contexts and seem to work differently based on
context (set_property can set multiple values, add_compile_options
can not/or is buggy)
* Could not call out to `sed` because cmake messed with the regular
expression before calling the external command. I did not see a way
around this.
* Makefile generators want to be single target. It wants a separate
directory for each target type. I saw some mentions on the web for
ways around this bug haven't look into it. The visual studio project
does support debug/release configurations in the same project (but
not unity/non-unity).
Calculate the number of file descriptors that are needed during
execution based on the configuration file, with a hard floor
of 1024, adjusting the limit if possible. Refuse to run if enough
fds are not available.
Additionally, allow administrators to limit the number of incoming
connections a configured port will accept. By default no limit is
imposed.
Replace Journal public data members with member function accessors
in order to make Journal lighter weight. The change makes a
Journal cheaper to pass by value.
Also add missing stream checks (e.g., calls to JLOG) to avoid
text processing that ultimately will not be stored in the log.
The Journal API is affected. There are two uses for the
Journal::Severity enum:
o It is used to declare a threshold which log messages must meet
in order to be logged.
o It declares the current logging level which will be compared
to the threshold.
Those uses that affect the threshold are now named threshold()
rather than severity() to make the uses easier to distinguish.
Additionally, Journal no longer carries a Severity variable.
All handling of the threshold() is now delegated to the
Journal::Sink.
Sinks are no longer constructed with a default threshold of
kWarning; their threshold must be passed in on construction.
Class io_list manages children that perform asynchronous
I/O operations. The treatment of close and destruction is
refactored to fix race conditions during exit.
