mysql-monitor_read_only_timeout is implemented as seconds count but documented as ms

[zahariash]

mysql-monitor_read_only_timeout seems to be implemented as seconds count (0cfb351) but documented as number of miliseconds (https://github.com/sysown/proxysql/blob/master/doc/monitor.md).

Config:

mysql> SHOW VARIABLES LIKE 'mysql-monitor_read%';
+----------------------------------+-------+
| Variable_name | Value |
+----------------------------------+-------+
| mysql-monitor_read_only_interval | 1500 |
| mysql-monitor_read_only_timeout | 500 |
+----------------------------------+-------+
2 rows in set (0.00 sec)

Logged timeout:

2016-12-12 05:27:01 MySQL_Monitor.cpp:612:monitor_read_only_thread(): [ERROR] Timeout on read_only check for localhost:3306 after 505838ms. If the server is overload, increase mysql-monitor_read_only_timeout. Assuming read_only=1

It's 500 seconds, which makes it probably useless...

[renecannao]

Thank you for the report.
What is incorrect is the error message.
The time unit returned from monotonic_time() is in microseconds, that is why mysql_thread___monitor_read_only_timeout (in milliseconds) is multiplied by 1000 in 0cfb351 .
I will fix the reported entry in the error log.

[zahariash]

I see, thanks.

[renecannao]

Reporting bug fixed in 1.3.1 (to be released today) and 1.4.0