Re: Finding the cause of a crashed table
Posted by:
Rick James
Date: February 06, 2015 01:43PM
Do the devices stay connected? Do they connect at the same time -- such as top of the hour? If so, could they be staggered?
500+ active connections could lead to resource contention, RAM and other. So, it may be safer to use
innodb_buffer_pool_size=1000M
instead of the 1500M I said before.
If 500 connections come in "simultaneously", MySQL will stumble over itself, leading to long latencies, spikes in CPU/IO, etc, but eventually work through all the requests.
Running out of ram leads to swapping, which is terribly bad for MySQL performance.
> Thanks for taking the time to look at my configuration.
You are welcome. However, because of the short Uptime, I only did half the analysis I could have.