locktree/tests/locktree_escalation_stalls.cc

/* -*- mode: C++; c-basic-offset: 4; indent-tabs-mode: nil -*- */
// vim: ft=cpp:expandtab:ts=8:sw=4:softtabstop=4:
#ident "$Id$"
/*======
This file is part of PerconaFT.


Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved.

    PerconaFT is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License, version 2,
    as published by the Free Software Foundation.

    PerconaFT is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with PerconaFT.  If not, see <http://www.gnu.org/licenses/>.

----------------------------------------

    PerconaFT is free software: you can redistribute it and/or modify
    it under the terms of the GNU Affero General Public License, version 3,
    as published by the Free Software Foundation.

    PerconaFT is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU Affero General Public License for more details.

    You should have received a copy of the GNU Affero General Public License
    along with PerconaFT.  If not, see <http://www.gnu.org/licenses/>.
======= */

#ident "Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved."

// This test verifies that small txn's do not get stalled for a long time by lock escalation.
// Two lock trees are used by the test: a big lock tree and a small lock tree.
// One big txn grabs lots of write locks on the big lock tree.
// Several small txn's grab a single write lock on the small lock tree.
// None of the locks conflict.
// Eventually, the locks for the big txn consume all of the lock tree memory, so lock escalation runs.
// The test measures the lock acquisition time and makes sure that the small txn's are not blocked for

// locktree_escalation_stalls -v --stalls 10
// verify that only big txn's get tagged with > 1 second stalls

#include <stdio.h>
#include "locktree.h"
#include "test.h"

using namespace toku;

static int verbose = 0;
static int killed = 0;

static void locktree_release_lock(locktree *lt, TXNID txn_id, int64_t left_k, int64_t right_k) {
    range_buffer buffer;
    buffer.create();
    DBT left; toku_fill_dbt(&left, &left_k, sizeof left_k);
    DBT right; toku_fill_dbt(&right, &right_k, sizeof right_k);
    buffer.append(&left, &right);
    lt->release_locks(txn_id, &buffer);
    buffer.destroy();
}

// grab a write range lock on int64 keys bounded by left_k and right_k
static int locktree_write_lock(locktree *lt, TXNID txn_id, int64_t left_k, int64_t right_k, bool big_txn) {
    DBT left; toku_fill_dbt(&left, &left_k, sizeof left_k);
    DBT right; toku_fill_dbt(&right, &right_k, sizeof right_k);
    return lt->acquire_write_lock(txn_id, &left, &right, nullptr, big_txn);
}

static void run_big_txn(locktree_manager *mgr UU(), locktree *lt, TXNID txn_id) {
    int64_t last_i = -1;
    for (int64_t i = 0; !killed; i++) {
        uint64_t t_start = toku_current_time_microsec();
        int r = locktree_write_lock(lt, txn_id, i, i, true);
        assert(r == 0);
        last_i = i;
        uint64_t t_end = toku_current_time_microsec();
        uint64_t t_duration = t_end - t_start;
        if (t_duration > 100000) {
            printf("%u %s %" PRId64 " %" PRIu64 "\n", toku_os_gettid(), __FUNCTION__, i, t_duration);
        }
        toku_pthread_yield();
    }
    if (last_i != -1)
        locktree_release_lock(lt, txn_id, 0, last_i); // release the range 0 .. last_i
}

static void run_small_txn(locktree_manager *mgr UU(), locktree *lt, TXNID txn_id, int64_t k) {
    for (int64_t i = 0; !killed; i++) {
        uint64_t t_start = toku_current_time_microsec();
        int r = locktree_write_lock(lt, txn_id, k, k, false);
        assert(r == 0);
        uint64_t t_end = toku_current_time_microsec();
        uint64_t t_duration = t_end - t_start;
        if (t_duration > 100000) {
            printf("%u %s %" PRId64 " %" PRIu64 "\n", toku_os_gettid(), __FUNCTION__, i, t_duration);
        }
        locktree_release_lock(lt, txn_id, k, k);
        toku_pthread_yield();
    }
}

struct arg {
    locktree_manager *mgr;
    locktree *lt;
    TXNID txn_id;
    int64_t k;
};

static void *big_f(void *_arg) {
    struct arg *arg = (struct arg *) _arg;
    run_big_txn(arg->mgr, arg->lt, arg->txn_id);
    return arg;
}

static void *small_f(void *_arg) {
    struct arg *arg = (struct arg *) _arg;
    run_small_txn(arg->mgr, arg->lt, arg->txn_id, arg->k);
    return arg;
}

static void e_callback(TXNID txnid, const locktree *lt, const range_buffer &buffer, void *extra) {
    if (verbose)
        printf("%u %s %" PRIu64 " %p %d %p\n", toku_os_gettid(), __FUNCTION__, txnid, lt, buffer.get_num_ranges(), extra);
}

static uint64_t get_escalation_count(locktree_manager &mgr) {
    LTM_STATUS_S ltm_status_test;
    mgr.get_status(&ltm_status_test);

    TOKU_ENGINE_STATUS_ROW key_status = NULL;
    // lookup keyname in status
    for (int i = 0; ; i++) {
        TOKU_ENGINE_STATUS_ROW status = &ltm_status_test.status[i];
        if (status->keyname == NULL)
            break;
        if (strcmp(status->keyname, "LTM_ESCALATION_COUNT") == 0) {
            key_status = status;
            break;
        }
    }
    assert(key_status);
    return key_status->value.num;
}

int main(int argc, const char *argv[]) {
    uint64_t stalls = 0;
    uint64_t max_lock_memory = 1000000000;
    for (int i = 1; i < argc; i++) {
        if (strcmp(argv[i], "-v") == 0 || strcmp(argv[i], "--verbose") == 0) {
            verbose++;
            continue;
        }
        if (strcmp(argv[i], "--stalls") == 0 && i+1 < argc) {
            stalls = atoll(argv[++i]);
            continue;
        }
        if (strcmp(argv[i], "--max_lock_memory") == 0 && i+1 < argc) {
            max_lock_memory = atoll(argv[++i]);
            continue;
        }
    }

    int r;

    // create a manager
    locktree_manager mgr;
    mgr.create(nullptr, nullptr, e_callback, nullptr);
    mgr.set_max_lock_memory(max_lock_memory);

    // create lock trees
    DICTIONARY_ID dict_id_0 = { .dictid = 1 };
    locktree *lt_0 = mgr.get_lt(dict_id_0, dbt_comparator, nullptr);

    DICTIONARY_ID dict_id_1 = { .dictid = 2 };
    locktree *lt_1 = mgr.get_lt(dict_id_1, dbt_comparator, nullptr);

    // create the worker threads
    struct arg big_arg = {&mgr, lt_0, 1000};
    pthread_t big_id;
    r = toku_pthread_create(
        toku_uninstrumented, &big_id, nullptr, big_f, &big_arg);
    assert(r == 0);

    const int n_small = 7;
    pthread_t small_ids[n_small];
    struct arg small_args[n_small];

    for (int i = 0; i < n_small; i++) {
        small_args[i] = {&mgr, lt_1, (TXNID)(2000 + i), i};
        r = toku_pthread_create(toku_uninstrumented,
                                &small_ids[i],
                                nullptr,
                                small_f,
                                &small_args[i]);
        assert(r == 0);
    }

    // wait for some escalations to occur
    while (get_escalation_count(mgr) < stalls) {
        sleep(1);
    }
    killed = 1;

    // cleanup
    void *ret;
    r = toku_pthread_join(big_id, &ret);
    assert(r == 0);

    for (int i = 0; i < n_small; i++) {
        r = toku_pthread_join(small_ids[i], &ret);
        assert(r == 0);
    }

    mgr.release_lt(lt_0);
    mgr.release_lt(lt_1);
    mgr.destroy();

    return 0;
}