dd50bd9bd4
system.batchlog will still have to be used while the cluster is upgrading from an older version, which doesn't know v2 yet. Re-add support for replaying v1 batchlogs. The switch to v2 will happen after the BATCHLOG_V2 cluster feature is enabled. The only external user -- storage_proxy -- only needs a minor adjustment: switch between the table names. The rest is handled transparently by the db/batchlog.hh interface and the batchlog_manager.
125 lines
3.4 KiB
C++
125 lines
3.4 KiB
C++
/*
|
|
* Copyright (C) 2015-present ScyllaDB
|
|
*
|
|
* Modified by ScyllaDB
|
|
*/
|
|
|
|
/*
|
|
* SPDX-License-Identifier: (LicenseRef-ScyllaDB-Source-Available-1.0 and Apache-2.0)
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <seastar/core/future.hh>
|
|
#include <seastar/core/sharded.hh>
|
|
#include <seastar/core/gate.hh>
|
|
#include <seastar/core/metrics_registration.hh>
|
|
#include <seastar/core/abort_source.hh>
|
|
|
|
#include "db_clock.hh"
|
|
|
|
#include <chrono>
|
|
#include <limits>
|
|
|
|
namespace cql3 {
|
|
|
|
class query_processor;
|
|
|
|
} // namespace cql3
|
|
|
|
namespace gms {
|
|
|
|
class feature_service;
|
|
|
|
} // namespace gms
|
|
|
|
namespace db {
|
|
|
|
class system_keyspace;
|
|
|
|
using all_batches_replayed = bool_class<struct all_batches_replayed_tag>;
|
|
|
|
struct batchlog_manager_config {
|
|
db_clock::duration replay_timeout;
|
|
uint64_t replay_rate = std::numeric_limits<uint64_t>::max();
|
|
std::chrono::milliseconds delay = std::chrono::milliseconds(0);
|
|
unsigned replay_cleanup_after_replays;
|
|
};
|
|
|
|
enum class batchlog_stage : int8_t {
|
|
initial,
|
|
failed_replay
|
|
};
|
|
|
|
class batchlog_manager : public peering_sharded_service<batchlog_manager> {
|
|
public:
|
|
using post_replay_cleanup = bool_class<class post_replay_cleanup_tag>;
|
|
|
|
struct stats {
|
|
uint64_t write_attempts = 0;
|
|
};
|
|
|
|
|
|
private:
|
|
static constexpr std::chrono::seconds replay_interval = std::chrono::seconds(60);
|
|
static constexpr uint32_t page_size = 128; // same as HHOM, for now, w/out using any heuristics. TODO: set based on avg batch size.
|
|
static constexpr std::chrono::seconds write_timeout = std::chrono::seconds(300);
|
|
|
|
using clock_type = lowres_clock;
|
|
|
|
stats _stats;
|
|
|
|
seastar::metrics::metric_groups _metrics;
|
|
|
|
cql3::query_processor& _qp;
|
|
db::system_keyspace& _sys_ks;
|
|
gms::feature_service& _fs;
|
|
db_clock::duration _replay_timeout;
|
|
uint64_t _replay_rate;
|
|
std::chrono::milliseconds _delay;
|
|
unsigned _replay_cleanup_after_replays = 100;
|
|
semaphore _sem{1};
|
|
seastar::named_gate _gate;
|
|
unsigned _cpu = 0;
|
|
seastar::abort_source _stop;
|
|
future<> _loop_done;
|
|
|
|
gc_clock::time_point _last_replay;
|
|
|
|
// Was the v1 -> v2 migration already done since last restart?
|
|
// The migration is attempted once after each restart. This is redundant but
|
|
// keeps thing simple. Once no upgrade path exists from a ScyllaDB version
|
|
// which can still produce v1 entries, this migration code can be removed.
|
|
bool _migration_done = false;
|
|
|
|
future<> maybe_migrate_v1_to_v2();
|
|
|
|
future<all_batches_replayed> replay_all_failed_batches_v1(post_replay_cleanup cleanup);
|
|
future<all_batches_replayed> replay_all_failed_batches_v2(post_replay_cleanup cleanup);
|
|
future<all_batches_replayed> replay_all_failed_batches(post_replay_cleanup cleanup);
|
|
public:
|
|
// Takes a QP, not a distributes. Because this object is supposed
|
|
// to be per shard and does no dispatching beyond delegating the the
|
|
// shard qp (which is what you feed here).
|
|
batchlog_manager(cql3::query_processor&, db::system_keyspace& sys_ks, gms::feature_service& fs, batchlog_manager_config config);
|
|
|
|
// abort the replay loop and return its future.
|
|
future<> drain();
|
|
future<> stop();
|
|
|
|
future<all_batches_replayed> do_batch_log_replay(post_replay_cleanup cleanup);
|
|
|
|
future<size_t> count_all_batches() const;
|
|
gc_clock::time_point get_last_replay() const {
|
|
return _last_replay;
|
|
}
|
|
|
|
const stats& get_stats() const {
|
|
return _stats;
|
|
}
|
|
private:
|
|
future<> batchlog_replay_loop();
|
|
};
|
|
|
|
}
|