Quote:
Originally Posted by zamler
Site was not responding for about 60 seconds a few minutes ago. My guess is the problem is not a network one but a CPU cycles one, meaning the processor is being overwhelmed (or the server is running out of memory) so request take a very long time to complete.
Mind you this is just a guess but I've seen very similar behaviour to what this forum is experiencing and that was the cause.
|
That's what's weird about it, when the pauses happen the CPUs are doing nothing, and the memory usage isn't anything different than normal, but the IO utilization for the database server is 100%. If I did some small IO tests at that time I'd get wildly varying throughput, but often in the KB/s rather than MB/s which is very very bad. Then things would smooth out and the disk IO would go down to its normal usage level and transfer rates would go back up to normal.
The machines are VMs and the disks are on a shared storage kind of thing so network problems made/make sense why the transfer rates would be wildly variable.