I’ve setup up a mysql percona cluster (pxc) deployment on a Rancher 2.6.3 (single node docker).
Everything is working but I’am having much lower performance than when provisioning pxc directly on baremetal, without using rancher/docker.
When I use it with rancher/docker container, doing some stress tests, I see that the disk iops are being much slower (30-50 MB/s) than when using without (more than 200 MB/s).
My worker nodes are baremetal servers with SSD’s and hardware RAID10, i am using bind-mounts.
I think this is related with some problem/limitation with the docker environment, but i can’t figure what might be.
Does anyone is experiencing this behavior?