Due to some quirky emergent behavior, the server can't really begin
synching until twice the default close time resolution of the genesis
ledger, which is 30 seconds, has passed. In effect, this causes a one
minute delay.
This commit adjusts the default close time resolution down to the
minimum allowed resoluion of 10 seconds, so the corresponding delay
is reduced by 67% down to 20 seconds. This should be enough time to
ensure the server has reasonable connectivity without unduly delaying
initial synch times.
This comment explains this patch and the associated patches
that should be folded into it. This paragraph should be removed
when the patches are folded after review.
This change significantly improves ledger sync and fetch
times while reducing memory consumption. The change affects
the code from that begins with SHAMap::getMissingNodes and runs
through to Database::threadEntry.
The existing code issues a number of async fetches which are then
handed off to the Database's pool of read threads to execute.
The results of each read are placed in the Database's positive
and negative caches. The caller waits for all reads to complete
and then retrieves the results out of these caches.
Among other issues, this means that the results of the first read
cannot be processed until the last read completes. Additionally,
all the results must sit in memory.
This patch changes the behavior so that each read operation has a
completion handler associated with it. The completion of the read
calls the handler, allowing the results of each read to be
processed as it completes. As this was the only reason the
negative and positive caches were needed, they can now be removed.
The read generation code is also no longer needed and is removed.
The batch fetch logic was never implemented or supported and is
removed.
gcc's implementation of `prm::synchronized_pool_resource` showed
extremely poor performance compared with
`boost::synchronized_pool_resouece`. Boost's implementation of pmr is
now used in all cases (previously it was only used when a standard
lib, like clang's, lacked an implementation of pmr).
This patch also makes a minor change where inner nodes are constructed
with sparse arrays, unless "dense" is explicitly requested.
Prior to this commit, the amendments that a server would vote in support
of or against could be configured both via the configuration file and
via the command line "feature" command. Changes made in the configuration
file would only be loaded once at server startup and changes made via the
command line take effect immediately but are not persisted across
restarts.
This commit deprecates management of amendments via the configuration
file and stores the relevant information in the `wallet.db` database
file.
1. On startup, the new code parses the configuration file.
2. If the `[veto_amendments]` or `[amendments]` sections are present,
we check if the `FeatureVotes` table is present in `wallet.db`.
3. If it is not, we create the `FeatureVotes` table and transfer the
settings from the config file.
4. Proceed normally but only reference the `FeatureVotes` table instead
of the config file.
5. Warns if the voting table already exists in `wallet.db` and there
exists voting sections in the config file. The config file is ignored
in this case.
This change addresses & closes#3366
* Found several functions called under lock that take a lock. Refactor
to require a lock as a parameter instead.
* Found several functions called under lock that don't take a lock, but
should. Refactored those as well to require a lock as a parameter.
Unit tests are counting test failures, process crashes, and process exit code
failures in the count. Since a failing tests causes the process exit code to
return failure, we get extra counts. This patch removes process exit code
failures from the count.
ReadViewFwdRange was storing a cached `end_` iterator that was lazily
created in an iterators `end()` function. When the cache is empty, and
the range it iterated from multiple threads, this creates a race
condition.
This change has performance consequences for "old style" for loops.
For example:
```
// don't do this
for(auto i = tx_range.begin(); i != tx_range.end(); ++i)
```
Can call the now expensive `end()` function more often than needed.
Range-based for loop (I.e. `for(auto const& t : tx_range)`) should be
used instead.
- Under some conditions, comparing `ReadViewFwdRange::iterators`
for equality could derefence an empty `std::unique_ptr` which
will result in a crash.
- Misuse of the `equal` API could result in a `std::bad_cast`
exception being thrown from when iterating transactions or
SLEs from the `OpenView`, `RawStateTable` and `Ledger` classes.
A large percentage of inner nodes only store a small number of children. Memory
can be saved by storing the inner node's children in sparse arrays. Measurements
show that on average a typical SHAMap's inner nodes can be stored using only 25%
of the original space.
This commit combines a number of cleanups, targeting both the
code structure and the code logic. Large changes include:
- Using more strongly-typed classes for SHAMap nodes, instead of relying
on runtime-time detection of class types. This change saves 16 bytes
of memory per node.
- Improving the interface of SHAMap::addGiveItem and SHAMap::addItem to
avoid the need for passing two bool arguments.
- Documenting the "copy-on-write" semantics that SHAMap uses to
efficiently track changes in individual nodes.
- Removing unused code and simplifying several APIs.
- Improving function naming.
- Simplify and consolidate code for parsing hex input.
- Replace beast::endian::order with boost::endian::order.
- Simplify CountedObject code.
- Remove pre-C++17 workarounds in favor of C++17 based solutions.
- Improve `base_uint` and simplify its hex-parsing interface by
consolidating the `SexHex` and `SetHexExact` methods into one
API: `parseHex` which forces callers to verify the result of
the operation; as a result some public-facing API endpoints
may now return errors when passed values that were previously
accepted.
- Remove the simple fallback implementations of SHA2 and RIPEMD
introduced to reduce our dependency on OpenSSL. The code is
slow and rarely, if ever, exercised and we rely on OpenSSL
functionality for Boost.ASIO as well.
- Provide separate functions for serializing depending on whether
one wants a "wire" version of a node, or one suitable for hashing.
- Remove unused functions
The existing SHAMapNodeID object has both a valid and an invalid state
and requirs callers to verify the state of an instance prior to using
it. A simple set of changes removes that restriction and ensures that
all instances are valid, making the code more robust.
This change also:
1. Introduces a new function to construct a SHAMapNodeID from a
serialized blob; and
2. Reduces the amount of constructors the class exposes.
- Limit the lifetime of a buffer that was only used in the early
phases of peer connection establishment but which lived on as
long as the peer was active.
- Cache the message used to transfer manifests, so it can be reused
instead of recreated for every peer connection.
- Improve the reading of partial messages by passing a hint to the
I/O layer if the number of bytes needed to complete the message
is known.
The existing code issues a PING to each peer every 8 seconds. While
frequent PINGs allow us to estimate a peer's latency with a high
degree of accuracy, this "inter-server polka dance" is inefficient
and not useful. This commit, if merged, reduces the PING frequency
to once every 60 seconds.
Additionally, this commit simplifies the PING handling logic and
merges the code used to check and disconnect peers which fail to
track the network directly into the timer callback.
When evaluating the fitness and usefulness of an outbound peer, the code
would incorrectly calculate the amount of time that the peer spent in
a non-useful state.
This commit, if merged, corrects the calculation and makes the timeout
values configurable by server operators.
Two new options are introduced in the 'overlay' stanza of the config
file. The default values, in seconds, are:
[overlay]
max_unknown_time = 600
max_diverged_time = 300
This commit replaces the `peers_max` configuration element which had
a predetermined split between incoming and outgoing connections with
two new configuration options, `peers_in_max` and `peers_out_max`,
which server operators can use to explicitly control the number of
incoming and outgoing peer slots.
There have been cases in the past where SFields have been defined
in such a way that they did not follow our conventions. In
particular, the string representation of an SField should match
the in-code name of the SField.
This change leverages the preprocessor to encourage SFields to
be properly constructed.
The suffixes of SField types are changed to be the same as
the suffixes of corresponding SerializedTypeIDs. This allows
The preprocessor to match types using simple name pasting.
Since the string representation of the SField is part of our
stable API, the name of sfPayChannel was changed to sfChannel.
This change allows sfChannel to follow our conventions while
making no changes to our external API.
* If multiple transactions are queued for the account, change the
account's sequence number in a temporary view before processing the
transaction.
* Adds a new "at()" interface to STObject which is identical to the
operator[], but easier to write and read when dealing with ptrs.
* Split the TxQ tests into two suites to speed up parallel run times.
This commit introduces a new configuration option that server
operators can set. The value is communicated to other servers
and is also reported via the `server_info` API.
The value is meant to allow third-party applications or tools
to group servers together. For example, a tool that visualizes
the network's topology can group servers together.
Similar to the "Domain" field in validator manifests, an operator
can claim any domain. Prior to relying on the value returned, the
domain should be verified by retrieving the xrp-ledger.toml file
from the domain and looking for the server's public key in the
`nodes` array.
* Jobs with no unit tests are counted as failures. Resolves#3474
* Crashed processes are counted as failures. Resolves#3600
* Any tests specified on the command line test do not have matching
suites are counted as failures.
* Remove unused CI manual test.
When processing the `tx` command, we will now load both the transaction
and its metadata directly from SQLite.
Previously the `tx` RPC call was querying SQLite for the transaction
and then separately querying the key-value store for the metadata.
