Just got a warning from a new budget shared hosting account telling me that sometimes the number of simultaneous MySQL requests exceeds it limit of 15 (at 3am local time for a tiny, tiny website that just sit there doing nothing 99% of the time). In all my years with concrete, I’ve never received this warning from any host, cheap or expensive. Is this something others have come up against? Is version 9 somehow more database intensive or is 15 just a ridiculously low limit? Any thoughts?
A wild guess. Could there be automated tasks running and opening simultaneous queries?
Good thought. I’m not a big user of automated tasks as in I’ve never set one up so probably not. Would these automated tasks show up in the history for each task? If so then no, there aren’t any because the last dates for any of the tasks was August 2023 when I last did them. It’s a very ‘brochure-like’ site where nothing much happens. For decades, I’ve been listening to hosting companies blaming clients for their problems and then the problem suddenly goes away because there was a problem on their server that they wouldn’t admit to so my gut says to do nothing and things will miraculously fix themselves.
15 isn’t a low number, I’d say that’s typical for shared hosting.
It should mean that you could have 15 simultaneous requests actively working and using the database in some way. But for modest sites you’re unlikely to have 15 requests all come in at once, as they may take only a fraction of a second to complete. So many actual visitors can effectively share those connections over a period of time.
(other factors start to come into play, like maximum processes, cpu limits, etc)
But abusive traffic, or some site indexes can hit websites pretty hard, with multiple requests all at once, and if those responses start to slow down, the connections can start to be used up.
I’ve actually seen DDOS style traffic that intentionally slows down communication to flood the number of connections a server can have. It’ll start a connection, then just not do http replies correctly, but the connections stay active for quite a while. In those cases I think the database connections would also be kept option as well.
So I’d look (or ask) about what traffic is hitting the site at that time in the morning. It could be just one or two abusive IP addresses causing the issue.
I asked them about AI bots hitting their servers because I understand that that’s a growing problem. They claim they put in safeguards a couple of weeks ago so my recent troubles shouldn’t be that caused by that but I’ll have a look at the server logs and see if I can see anything.
At the time of the outage, the logs show that the site was just sitting there waiting for my website monitor service to hit the Home page once every 5 minutes. I’m calling this one busted.