Good morning!
My next issue I need to struggle with is weird. It is even hard to describe...
- I have my 3-instances proxysql cluster working with 3-nodes mysql cluster. Mysql instances are using simple replication (not even a group-replication), but that is not important here.
- I have some rules, I have two hostgroups (read (20)/write (10)) and that works just fine.
- I'm using default ports - 6032 and 6033.
- All communication on 6033 (mysql backend) work as normal.
So, to monitor my connections, to have some insight into the traffic I created a few "one-liner" scripts to fetch data from ProxySQL. Like:
"select sum(sum_time) as total_time from stats_mysql_query_digest where username = 'xxx' and hostgroup = '10'
And the 'scripts' like that - around 10 already - are triggered every 2 minutes (by the third party software - Zabbix in my case)
And now there is my problem. Every day, around the same time of a day, those scripts start to getting stuck with no result from Proxy Admin Interface. It looks like Admin stopped responding and handling those requests. When it happens I cannot - of course - even connect to Admin from the command line. Every new run is added to the stuck and waits.
Here is the list of stacked queries checked during a few seconds. It is clearly see that this number grows and now it is around 8 thousands. But, we can see that it also goes down a bit sometimes (8269 -> 8268 -> 8264 -> 8257 -> after few seconds it went up to 8323)
Again, during that time everything is fine with the proxy -> mysql connection. All requests are handled on the fly.
The way for me to clear that stack is to restart entire proxysql. And then I have another day when all my scripts are handled in a second.
Anyone saw something like this before? Any clue how to troubleshoot it?
Thanks a lot!
Sławek