ARCHIVE/clio
1
0
Fork 0
mirror of https://github.com/XRPLF/clio.git synced 2025-11-10 14:55:50 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
69f5025a29e164a948b47ff2658be96d68096a39
clio/src/etl/impl/CacheLoader.h
Alex Kremer 0eaaa1fb31 Add workaround for async_compose (#841) 
Fixes #840
2023-09-18 18:52:32 +01:00

436 lines
16 KiB
C++
Raw Blame History

//------------------------------------------------------------------------------
/*
This file is part of clio: https://github.com/XRPLF/clio
Copyright (c) 2023, the clio developers.
Permission to use, copy, modify, and distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.
THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
//==============================================================================
#pragma once
#include <data/BackendInterface.h>
#include <util/log/Logger.h>
#include <ripple/proto/org/xrpl/rpc/v1/xrp_ledger.grpc.pb.h>
#include <boost/algorithm/string.hpp>
#include <boost/asio/spawn.hpp>
#include <boost/beast/core.hpp>
#include <boost/beast/core/string.hpp>
#include <boost/beast/websocket.hpp>
#include <grpcpp/grpcpp.h>
#include <chrono>
#include <mutex>
#include <thread>
namespace etl::detail {
/**
* @brief Cache loading interface
*/
template <typename CacheType>
class CacheLoader
{
enum class LoadStyle { ASYNC, SYNC, NOT_AT_ALL };
util::Logger log_{"ETL"};
std::reference_wrapper<boost::asio::io_context> ioContext_;
std::shared_ptr<BackendInterface> backend_;
std::reference_wrapper<CacheType> cache_;
LoadStyle cacheLoadStyle_ = LoadStyle::ASYNC;
// number of diffs to use to generate cursors to traverse the ledger in parallel during initial cache download
size_t numCacheDiffs_ = 32;
// number of markers to use at one time to traverse the ledger in parallel during initial cache download
size_t numCacheMarkers_ = 48;
// number of ledger objects to fetch concurrently per marker during cache download
size_t cachePageFetchSize_ = 512;
struct ClioPeer
{
std::string ip;
int port;
};
std::vector<ClioPeer> clioPeers_;
std::thread thread_;
std::atomic_bool stopping_ = false;
public:
CacheLoader(
util::Config const& config,
boost::asio::io_context& ioc,
std::shared_ptr<BackendInterface> const& backend,
CacheType& ledgerCache)
: ioContext_{std::ref(ioc)}, backend_{backend}, cache_{ledgerCache}
{
if (config.contains("cache"))
{
auto const cache = config.section("cache");
if (auto entry = cache.maybeValue<std::string>("load"); entry)
{
if (boost::iequals(*entry, "sync"))
cacheLoadStyle_ = LoadStyle::SYNC;
if (boost::iequals(*entry, "async"))
cacheLoadStyle_ = LoadStyle::ASYNC;
if (boost::iequals(*entry, "none") or boost::iequals(*entry, "no"))
cacheLoadStyle_ = LoadStyle::NOT_AT_ALL;
}
numCacheDiffs_ = cache.valueOr<size_t>("num_diffs", numCacheDiffs_);
numCacheMarkers_ = cache.valueOr<size_t>("num_markers", numCacheMarkers_);
cachePageFetchSize_ = cache.valueOr<size_t>("page_fetch_size", cachePageFetchSize_);
if (auto peers = cache.maybeArray("peers"); peers)
{
for (auto const& peer : *peers)
{
auto ip = peer.value<std::string>("ip");
auto port = peer.value<uint32_t>("port");
// todo: use emplace_back when clang is ready
clioPeers_.push_back({ip, port});
}
unsigned seed = std::chrono::system_clock::now().time_since_epoch().count();
std::shuffle(std::begin(clioPeers_), std::end(clioPeers_), std::default_random_engine(seed));
}
}
}
~CacheLoader()
{
stop();
if (thread_.joinable())
thread_.join();
}
/**
* @brief Populates the cache by walking through the given ledger.
*
* Should only be called once. The default behavior is to return immediately and populate the cache in the
* background. This can be overridden via config parameter, to populate synchronously, or not at all.
*/
void
load(uint32_t seq)
{
if (cacheLoadStyle_ == LoadStyle::NOT_AT_ALL)
{
cache_.get().setDisabled();
LOG(log_.warn()) << "Cache is disabled. Not loading";
return;
}
if (cache_.get().isFull())
{
assert(false);
return;
}
if (clioPeers_.size() > 0)
{
boost::asio::spawn(ioContext_.get(), [this, seq](boost::asio::yield_context yield) {
for (auto const& peer : clioPeers_)
{
// returns true on success
if (loadCacheFromClioPeer(seq, peer.ip, std::to_string(peer.port), yield))
return;
}
// if we couldn't successfully load from any peers, load from db
loadCacheFromDb(seq);
});
return;
}
else
{
loadCacheFromDb(seq);
}
// If loading synchronously, poll cache until full
while (cacheLoadStyle_ == LoadStyle::SYNC && not cache_.get().isFull())
{
LOG(log_.debug()) << "Cache not full. Cache size = " << cache_.get().size() << ". Sleeping ...";
std::this_thread::sleep_for(std::chrono::seconds(10));
if (cache_.get().isFull())
LOG(log_.info()) << "Cache is full. Cache size = " << cache_.get().size();
}
}
void
stop()
{
stopping_ = true;
}
private:
bool
loadCacheFromClioPeer(
uint32_t ledgerIndex,
std::string const& ip,
std::string const& port,
boost::asio::yield_context yield)
{
LOG(log_.info()) << "Loading cache from peer. ip = " << ip << " . port = " << port;
namespace beast = boost::beast; // from <boost/beast.hpp>
namespace http = beast::http; // from <boost/beast/http.hpp>
namespace websocket = beast::websocket; // from
namespace net = boost::asio; // from
using tcp = boost::asio::ip::tcp; // from
try
{
boost::beast::error_code ec;
// These objects perform our I/O
tcp::resolver resolver{ioContext_.get()};
LOG(log_.trace()) << "Creating websocket";
auto ws = std::make_unique<websocket::stream<beast::tcp_stream>>(ioContext_.get());
// Look up the domain name
auto const results = resolver.async_resolve(ip, port, yield[ec]);
if (ec)
return {};
LOG(log_.trace()) << "Connecting websocket";
// Make the connection on the IP address we get from a lookup
ws->next_layer().async_connect(results, yield[ec]);
if (ec)
return false;
LOG(log_.trace()) << "Performing websocket handshake";
// Perform the websocket handshake
ws->async_handshake(ip, "/", yield[ec]);
if (ec)
return false;
std::optional<boost::json::value> marker;
LOG(log_.trace()) << "Sending request";
auto getRequest = [&](auto marker) {
boost::json::object request = {
{"command", "ledger_data"},
{"ledger_index", ledgerIndex},
{"binary", true},
{"out_of_order", true},
{"limit", 2048}};
if (marker)
request["marker"] = *marker;
return request;
};
bool started = false;
size_t numAttempts = 0;
do
{
// Send the message
ws->async_write(net::buffer(boost::json::serialize(getRequest(marker))), yield[ec]);
if (ec)
{
LOG(log_.error()) << "error writing = " << ec.message();
return false;
}
beast::flat_buffer buffer;
ws->async_read(buffer, yield[ec]);
if (ec)
{
LOG(log_.error()) << "error reading = " << ec.message();
return false;
}
auto raw = beast::buffers_to_string(buffer.data());
auto parsed = boost::json::parse(raw);
if (!parsed.is_object())
{
LOG(log_.error()) << "Error parsing response: " << raw;
return false;
}
LOG(log_.trace()) << "Successfully parsed response " << parsed;
if (auto const& response = parsed.as_object(); response.contains("error"))
{
LOG(log_.error()) << "Response contains error: " << response;
auto const& err = response.at("error");
if (err.is_string() && err.as_string() == "lgrNotFound")
{
++numAttempts;
if (numAttempts >= 5)
{
LOG(log_.error()) << " ledger not found at peer after 5 attempts. "
"peer = "
<< ip << " ledger = " << ledgerIndex
<< ". Check your config and the health of the peer";
return false;
}
LOG(log_.warn()) << "Ledger not found. ledger = " << ledgerIndex
<< ". Sleeping and trying again";
std::this_thread::sleep_for(std::chrono::seconds(1));
continue;
}
return false;
}
started = true;
auto const& response = parsed.as_object()["result"].as_object();
if (!response.contains("cache_full") || !response.at("cache_full").as_bool())
{
LOG(log_.error()) << "cache not full for clio node. ip = " << ip;
return false;
}
if (response.contains("marker"))
marker = response.at("marker");
else
marker = {};
auto const& state = response.at("state").as_array();
std::vector<data::LedgerObject> objects;
objects.reserve(state.size());
for (auto const& ledgerObject : state)
{
auto const& obj = ledgerObject.as_object();
data::LedgerObject stateObject = {};
if (!stateObject.key.parseHex(obj.at("index").as_string().c_str()))
{
LOG(log_.error()) << "failed to parse object id";
return false;
}
boost::algorithm::unhex(obj.at("data").as_string().c_str(), std::back_inserter(stateObject.blob));
objects.push_back(std::move(stateObject));
}
cache_.get().update(objects, ledgerIndex, true);
if (marker)
LOG(log_.debug()) << "At marker " << *marker;
} while (marker || !started);
LOG(log_.info()) << "Finished downloading ledger from clio node. ip = " << ip;
cache_.get().setFull();
return true;
}
catch (std::exception const& e)
{
LOG(log_.error()) << "Encountered exception : " << e.what() << " - ip = " << ip;
return false;
}
}
void
loadCacheFromDb(uint32_t seq)
{
std::vector<data::LedgerObject> diff;
std::vector<std::optional<ripple::uint256>> cursors;
auto append = [](auto&& a, auto&& b) { a.insert(std::end(a), std::begin(b), std::end(b)); };
for (size_t i = 0; i < numCacheDiffs_; ++i)
{
append(diff, data::synchronousAndRetryOnTimeout([&](auto yield) {
return backend_->fetchLedgerDiff(seq - i, yield);
}));
}
std::sort(diff.begin(), diff.end(), [](auto a, auto b) {
return a.key < b.key || (a.key == b.key && a.blob.size() < b.blob.size());
});
diff.erase(std::unique(diff.begin(), diff.end(), [](auto a, auto b) { return a.key == b.key; }), diff.end());
cursors.push_back({});
for (auto const& obj : diff)
if (obj.blob.size())
cursors.push_back({obj.key});
cursors.push_back({});
std::stringstream cursorStr;
for (auto const& c : cursors)
if (c)
cursorStr << ripple::strHex(*c) << ", ";
LOG(log_.info()) << "Loading cache. num cursors = " << cursors.size() - 1;
LOG(log_.trace()) << "cursors = " << cursorStr.str();
thread_ = std::thread{[this, seq, cursors = std::move(cursors)]() {
auto startTime = std::chrono::system_clock::now();
auto markers = std::make_shared<std::atomic_int>(0);
auto numRemaining = std::make_shared<std::atomic_int>(cursors.size() - 1);
for (size_t i = 0; i < cursors.size() - 1; ++i)
{
auto const start = cursors.at(i);
auto const end = cursors.at(i + 1);
markers->wait(numCacheMarkers_);
++(*markers);
boost::asio::spawn(
ioContext_.get(),
[this, seq, start, end, numRemaining, startTime, markers](boost::asio::yield_context yield) {
auto cursor = start;
std::string cursorStr =
cursor.has_value() ? ripple::strHex(cursor.value()) : ripple::strHex(data::firstKey);
LOG(log_.debug()) << "Starting a cursor: " << cursorStr << " markers = " << *markers;
while (not stopping_)
{
auto res = data::retryOnTimeout([this, seq, &cursor, yield]() {
return backend_->fetchLedgerPage(cursor, seq, cachePageFetchSize_, false, yield);
});
cache_.get().update(res.objects, seq, true);
if (!res.cursor || (end && *(res.cursor) > *end))
break;
LOG(log_.trace()) << "Loading cache. cache size = " << cache_.get().size()
<< " - cursor = " << ripple::strHex(res.cursor.value())
<< " start = " << cursorStr << " markers = " << *markers;
cursor = std::move(res.cursor);
}
--(*markers);
markers->notify_one();
if (--(*numRemaining) == 0)
{
auto endTime = std::chrono::system_clock::now();
auto duration = std::chrono::duration_cast<std::chrono::seconds>(endTime - startTime);
LOG(log_.info()) << "Finished loading cache. cache size = " << cache_.get().size()
<< ". Took " << duration.count() << " seconds";
cache_.get().setFull();
}
else
{
LOG(log_.info()) << "Finished a cursor. num remaining = " << *numRemaining
<< " start = " << cursorStr << " markers = " << *markers;
}
});
}
}};
}
};
} // namespace etl::detail
Reference in New Issue View Git Blame Copy Permalink