我们通过源代码,找到下面一段,该段实现了上述日志的输出。
if ((my_now – rli->mts_last_online_stat)>=
mts_online_stat_period)
{
sql_print_information(“Multi-threadedslave statistics%s: “
“seconds elapsed = %lu; “
“events assigned = %llu; “
“worker queues filled over overrun level = %lu;”
“waited due a Worker queue full = %lu; “
“waited due the total size = %lu; “
“waited at clock conflicts = %llu “
“waited(count) when Workers occupied = %lu “
“waited when Workers occupied = %llu”,
rli->get_for_channel_str(),
static_cast<unsignedlong>
(my_now – rli->mts_last_online_stat),
rli->mts_events_assigned,
rli->mts_wq_overrun_cnt,
rli->mts_wq_overfill_cnt,
rli->wq_size_waits_cnt,
rli->mts_total_wait_overlap,
rli->mts_wq_no_underrun_cnt,
rli->mts_total_wait_worker_avail);
rli->mts_last_online_stat=my_now;
通过这一句(my_now – rli->mts_last_online_stat), 以及最后一句rli->mts_last_online_stat=my_now; 可以得知, seconds elapsed 就是上一次统计跟这一次统计的时间间隔。
而mts_online_stat_period =120秒,硬代码,因此就是几乎每隔120秒,就有上述日志的输出。 通过进一步查看原代码,初步了解上述日志信息的含义,如下:
events assigned:总共有多少个event被分配执行,计的是总数。
worker queues filled over overrun level:多线程同步中,worker 的私有队列长度超长的次数,计的是总数。
waited due a Worker queue full :因为worker的队列超长而产生等待的次数,计的是总数。
waited due the total size :超过最大size的次数,这个由参数slave_pending_jobs_size_max 指定。
waited at clock conflicts :因为逻辑时间产生冲突的等待时间,单位是纳秒。
waited (count) when Workers occupied :因为workder被占用而出现等待的次数。(总计值)。
waited when Workersoccupied :因为workder被占用而出现等待的总时间,总计值,单位是纳秒。
第三种:page_cleaner线程的输出日志
如图,信息如下:
2016-03-24T17:45:28.005117Z 0 [Note] InnoDB:page_cleaner: 1000ms intended loop took 4750ms.The settings might not beoptimal. (flushed=1519 and evicted=0, during the time.)
查找源代码,发现是上面的日志由下面一段代码输出:
if (ret_sleep == OS_SYNC_TIME_EXCEEDED) {
ulint curr_time = ut_time_ms();
if (curr_time >next_loop_time + 3000) {
if (warn_count == 0) {
ib::info() << “page_cleaner: 1000ms”
” intended loop took “
<<1000 + curr_time
– next_loop_time
<<“ms. The settings might not”
” be optimal. (flushed=”
<<n_flushed_last
<<” and evicted=”
<<n_evicted
<<“, during the time.)”;
if (warn_interval >300) {
warn_interval= 600;
}else {
warn_interval*= 2;
}
warn_count= warn_interval;
} else {
–warn_count;
}
} else {
/* reset counter */
warn_interval= 1;
warn_count= 0;
}
next_loop_time= curr_time + 1000;
n_flushed_last= n_evicted = 0;
}
通过分析源代码, 输出上述日志的条件是curr_time> next_loop_time + 3000 ,即比原定的循环时间next_loop_time多3000毫秒,而next_loop_time的标准时间是1000毫秒,即1秒钟做一次刷新页的操作。
loop took 4750ms ,即是这次刷新循环的实际经历时间。
后面还有一个(flushed=1519 and evicted=0,during the time.)这样的日志,即对应n_flushed_last与n_evicted 变量,而这两个变量又由n_flushed_list与n_flushed_lru赋值。
./storage/innobase/buf/:3322: n_flushed_last +=n_flushed_list;
./storage/innobase/buf/:3321: n_evicted += n_flushed_lru;
而n_flushed_list与n_flushed_lru赋值函数为pc_wait_finished,如下,我们来看看该函数的注释。
pc_wait_finished(&n_flushed_lru,&n_flushed_list);
/**
Wait until all flush requests are finished.
@param n_flushed_lru numberof pages flushed from the end of the LRU list.
@param n_flushed_list numberof pages flushed from the end of the
flush_list.
@return trueif all flush_list flushing batch were success. */
static
bool
pc_wait_finished(
ulint* n_flushed_lru,
ulint* n_flushed_list)
{
。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。
}
通过源代码的注释,我们获知如下信息:
n_flushed_lru number of pages flushed from the end of the LRU list.
n_flushed_lru 这个值表示从lru 列表尾部刷新的页数,也就是日志中如evicted=0 指标的所表示的值,如果该值不为零,则存在innodb buffer不够的嫌疑。
n_flushed_list 这个是从刷新列表中刷新的页数,也就是脏页数,也就是日志中flushed=1519 的值。