ConnectionPoolWithFailover.cpp 7.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
#include <Client/ConnectionPoolWithFailover.h>

#include <Poco/Net/NetException.h>
#include <Poco/Net/DNS.h>

#include <Common/getFQDNOrHostName.h>
#include <Common/ProfileEvents.h>
#include <Interpreters/Settings.h>


namespace ProfileEvents
{
    extern const Event DistributedConnectionStaleReplica;
}

namespace DB
{

ConnectionPoolWithFailover::ConnectionPoolWithFailover(
20
        ConnectionPoolPtrs nested_pools_,
21 22 23
        LoadBalancing load_balancing,
        size_t max_tries_,
        time_t decrease_error_period_)
24
    : Base(std::move(nested_pools_), max_tries_, decrease_error_period_, &Logger::get("ConnectionPoolWithFailover"))
25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
    , default_load_balancing(load_balancing)
{
    const std::string & local_hostname = getFQDNOrHostName();

    hostname_differences.resize(nested_pools.size());
    for (size_t i = 0; i < nested_pools.size(); ++i)
    {
        ConnectionPool & connection_pool = dynamic_cast<ConnectionPool &>(*nested_pools[i]);
        const std::string & host = connection_pool.getHost();

        size_t hostname_difference = 0;
        for (size_t i = 0; i < std::min(local_hostname.length(), host.length()); ++i)
            if (local_hostname[i] != host[i])
                ++hostname_difference;

        hostname_differences[i] = hostname_difference;
    }
}

44
IConnectionPool::Entry ConnectionPoolWithFailover::get(const Settings * settings)
45 46 47 48 49 50
{
    Base::TryGetEntryFunc try_get_entry = [&](NestedPool & pool, std::string & fail_message)
    {
        return tryGetEntry(pool, fail_message, settings);
    };

51 52 53 54 55 56 57 58 59 60 61 62 63 64
    Base::GetPriorityFunc get_priority;
    switch (settings ? LoadBalancing(settings->load_balancing) : default_load_balancing)
    {
    case LoadBalancing::NEAREST_HOSTNAME:
        get_priority = [&](size_t i) { return hostname_differences[i]; };
        break;
    case LoadBalancing::IN_ORDER:
        get_priority = [](size_t i) { return i; };
        break;
    case LoadBalancing::RANDOM:
        break;
    }

    return Base::get(try_get_entry, get_priority);
65 66
}

67
std::vector<IConnectionPool::Entry> ConnectionPoolWithFailover::getMany(const Settings * settings, PoolMode pool_mode)
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
{
    Base::TryGetEntryFunc try_get_entry = [&](NestedPool & pool, std::string & fail_message)
    {
        return tryGetEntry(pool, fail_message, settings);
    };
    return getManyImpl(settings, pool_mode, try_get_entry);
}

std::vector<IConnectionPool::Entry> ConnectionPoolWithFailover::getManyChecked(
        const Settings * settings, PoolMode pool_mode, const QualifiedTableName & table_to_check)
{
    Base::TryGetEntryFunc try_get_entry = [&](NestedPool & pool, std::string & fail_message)
    {
        return tryGetEntry(pool, fail_message, settings, &table_to_check);
    };
    return getManyImpl(settings, pool_mode, try_get_entry);
}

std::vector<ConnectionPool::Entry> ConnectionPoolWithFailover::getManyImpl(
        const Settings * settings,
        PoolMode pool_mode,
        const Base::TryGetEntryFunc & try_get_entry)
{
    size_t min_entries = (settings && settings->skip_unavailable_shards) ? 0 : 1;
    size_t max_entries;
    if (pool_mode == PoolMode::GET_ALL)
    {
        min_entries = nested_pools.size();
        max_entries = nested_pools.size();
    }
    else if (pool_mode == PoolMode::GET_ONE)
        max_entries = 1;
    else if (pool_mode == PoolMode::GET_MANY)
        max_entries = settings ? size_t(settings->max_parallel_replicas) : 1;
    else
        throw DB::Exception("Unknown pool allocation mode", DB::ErrorCodes::LOGICAL_ERROR);

    Base::GetPriorityFunc get_priority;
106
    switch (settings ? LoadBalancing(settings->load_balancing) : default_load_balancing)
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
    {
    case LoadBalancing::NEAREST_HOSTNAME:
        get_priority = [&](size_t i) { return hostname_differences[i]; };
        break;
    case LoadBalancing::IN_ORDER:
        get_priority = [](size_t i) { return i; };
        break;
    case LoadBalancing::RANDOM:
        break;
    }

    bool fallback_to_stale_replicas = settings ? bool(settings->fallback_to_stale_replicas_for_distributed_queries) : true;

    return Base::getMany(min_entries, max_entries, try_get_entry, get_priority, fallback_to_stale_replicas);
}

ConnectionPoolWithFailover::Base::TryResult
ConnectionPoolWithFailover::tryGetEntry(
        IConnectionPool & pool,
        std::string & fail_message,
        const Settings * settings,
        const QualifiedTableName * table_to_check)
{
    TryResult result;
    try
    {
        result.entry = pool.get(settings);

        UInt64 max_allowed_delay = settings ? UInt64(settings->max_replica_delay_for_distributed_queries) : 0;

        String server_name;
        UInt64 server_version_major;
        UInt64 server_version_minor;
        UInt64 server_revision;
        if (table_to_check && max_allowed_delay)
            result.entry->getServerVersion(server_name, server_version_major, server_version_minor, server_revision);

        if (!table_to_check || !max_allowed_delay || server_revision < DBMS_MIN_REVISION_WITH_TABLES_STATUS)
        {
            result.entry->forceConnected();
147
            result.is_up_to_date = true;
148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
            return result;
        }

        /// Only the delay of the remote table of the corresponding Distributed table is taken into account.
        /// TODO: calculate delay for joined tables also.
        TablesStatusRequest status_request;
        status_request.tables = { *table_to_check };

        auto status_response = result.entry->getTablesStatus(status_request);
        if (status_response.table_states_by_id.size() != status_request.tables.size())
            throw Exception(
                    "Bad TablesStatus response (from " + result.entry->getDescription() + ")",
                    ErrorCodes::LOGICAL_ERROR);

        UInt32 max_delay = 0;
        for (const auto & kv: status_response.table_states_by_id)
        {
            const TableStatus & status = kv.second;

            if (status.is_replicated)
                max_delay = std::max(max_delay, status.absolute_delay);
        }

171 172 173
        if (max_delay < max_allowed_delay)
            result.is_up_to_date = true;
        else
174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
        {
            result.is_up_to_date = false;
            result.staleness = max_delay;

            LOG_WARNING(
                    log, "Connection " << result.entry->getDescription() << " has unacceptable replica delay "
                    << "for table " << table_to_check->database << "." << table_to_check->table
                    << ": "  << max_delay);
            ProfileEvents::increment(ProfileEvents::DistributedConnectionStaleReplica);
        }
    }
    catch (const Exception & e)
    {
        if (e.code() != ErrorCodes::NETWORK_ERROR && e.code() != ErrorCodes::SOCKET_TIMEOUT
            && e.code() != ErrorCodes::ATTEMPT_TO_READ_AFTER_EOF)
            throw;

        fail_message = "Code: " + std::to_string(e.code()) + ", e.displayText() = " + e.displayText() + ", e.what() = " + e.what();

        if (!result.entry.isNull())
        {
            result.entry->disconnect();
            result.entry = Entry();
        }
    }
    return result;
};

}