Files
scylladb/replica/distributed_loader.hh
Petr Gusev 5a5d664a5a init_system_keyspace: refactoring towards explicit load phases
We aim (#12642) to use the schema commit log
for raft tables. Now they are loaded at
the first call to init_system_keyspace in
main.cc, but the schema commitlog is only
initialized shortly before the second
call. This is important, since the schema
commitlog initialization
(database::before_schema_keyspace_init)
needs to access schema commitlog feature,
which is loaded from system.scylla_local
and therefore is only available after the
first init_system_keyspace call.

So the idea is to defer the loading of the raft tables
until the second call to init_system_keyspace,
just as it works for schema tables.
For this we need a tool to mark which tables
should be loaded in the first or second phase.

To do this, in this patch we introduce system_table_load_phase
enum. It's set in the schema_static_props for schema tables.
It replaces the system_keyspace::table_selector in the
signature of init_system_keyspace.

The call site for populate_keyspace in init_system_keyspace
was changed, table_selector.contains_keyspace was replaced with
db.local().has_keyspace. This check prevents calling
populate_keyspace(system_schema) on phase1, but allows for
populate_keyspace(system) on phase2 (to init raft tables).
On this second call some tables from system keyspace
(e.g. system.local) may have already been populated on phase1.
This check protects from double-populating them, since every
populated cf is marked as ready_for_writes.
2023-03-24 15:54:46 +04:00

108 lines
3.9 KiB
C++

/*
* Copyright (C) 2018-present ScyllaDB
*/
/*
* SPDX-License-Identifier: AGPL-3.0-or-later
*/
#pragma once
#include <seastar/core/future.hh>
#include <seastar/core/distributed.hh>
#include <seastar/core/sstring.hh>
#include <seastar/core/file.hh>
#include <seastar/util/bool_class.hh>
#include <vector>
#include <functional>
#include <filesystem>
#include "seastarx.hh"
#include "compaction/compaction_descriptor.hh"
#include "db/system_keyspace.hh"
#include "sstables/sstable_directory.hh"
namespace replica {
class database;
class table;
using column_family = table;
}
namespace db {
class config;
class system_distributed_keyspace;
class system_keyspace;
namespace view {
class view_update_generator;
}
}
namespace sstables {
class entry_descriptor;
class foreign_sstable_open_info;
}
namespace service {
class storage_proxy;
class storage_service;
class raft_group_registry;
}
namespace gms {
class gossiper;
}
class distributed_loader_for_tests;
namespace replica {
class table_populator;
class distributed_loader {
friend class ::distributed_loader_for_tests;
friend class table_populator;
static future<> reshape(sharded<sstables::sstable_directory>& dir, sharded<replica::database>& db, sstables::reshape_mode mode,
sstring ks_name, sstring table_name, sstables::compaction_sstable_creator_fn creator, std::function<bool (const sstables::shared_sstable&)> filter, io_priority_class iop);
static future<> reshard(sharded<sstables::sstable_directory>& dir, sharded<replica::database>& db, sstring ks_name, sstring table_name, sstables::compaction_sstable_creator_fn creator,
io_priority_class iop);
static future<> process_sstable_dir(sharded<sstables::sstable_directory>& dir, sstables::sstable_directory::process_flags flags);
static future<> lock_table(sharded<sstables::sstable_directory>& dir, sharded<replica::database>& db, sstring ks_name, sstring cf_name);
static future<size_t> make_sstables_available(sstables::sstable_directory& dir,
sharded<replica::database>& db, sharded<db::view::view_update_generator>& view_update_generator,
bool needs_view_update, sstring ks, sstring cf);
static future<> populate_keyspace(distributed<replica::database>& db, sstring datadir, sstring ks_name);
static future<> cleanup_column_family_temp_sst_dirs(sstring sstdir);
static future<> handle_sstables_pending_delete(sstring pending_deletes_dir);
public:
static future<> init_system_keyspace(sharded<db::system_keyspace>& sys_ks, distributed<replica::database>& db, distributed<service::storage_service>& ss, sharded<gms::gossiper>& g, sharded<service::raft_group_registry>& raft_gr, db::config& cfg, system_table_load_phase phase);
static future<> init_non_system_keyspaces(distributed<replica::database>& db, distributed<service::storage_proxy>& proxy, sharded<db::system_keyspace>& sys_ks);
/**
* Marks a keyspace (by name) as "prioritized" on bootstrap.
* This will effectively let it bypass concurrency control.
* The only real use for this is to avoid certain chicken and
* egg issues.
*
* May only be called pre-bootstrap on main shard.
* Required for enterprise. Do _not_ remove.
*/
static void mark_keyspace_as_load_prio(const sstring&);
// Scan sstables under upload directory. Return a vector with smp::count entries.
// Each entry with index of idx should be accessed on shard idx only.
// Each entry contains a vector of sstables for this shard.
// The table UUID is returned too.
static future<std::tuple<table_id, std::vector<std::vector<sstables::shared_sstable>>>>
get_sstables_from_upload_dir(distributed<replica::database>& db, sstring ks, sstring cf);
static future<> process_upload_dir(distributed<replica::database>& db, distributed<db::system_distributed_keyspace>& sys_dist_ks,
distributed<db::view::view_update_generator>& view_update_generator, sstring ks_name, sstring cf_name);
};
}