I'm messing with Proxmox for the first time and experiencing some behavior that I can't seem to find a good reason for.
I've got a process that loads data into a mysql database - it's essentially several gigs worth of batched inserts. It runs pretty consistently in each environment.
These are the three environments I'm comparing:
1. MySql in a Docker container on Ubuntu, bare metal, with the Docker volume on a local SATA drive.
- This is the machine I've been running this process on for years.
- The process takes about an hour.
2. MySql in a Docker container on an Ubuntu Cloud Image Proxmox VM with the VM image on a Synology NFS, 1GbE.
- The NFS just has a single, old, SATA drive, because I wanted to test the network impact on the process.
- The process runs about 10% slower than the original setup, which seems like a reasonable trade-off to me, especially on 1GbE.
3. MySql on an Ubuntu Cloud Image Proxmox VM with the VM image on the machine's NVMe.
- I've tried this setup with MySql both directly on Ubuntu, and in a Docker container
- fio benchmarks about 10x better on this VM than the VM on NFS, which is understandable.
- The process takes about 100 minutes, which seems really strange, especially considering how well the NFS setup did.
Scenarios 2 and 3 are the same machine, different VMs.
I've only tried the VirtIO SCSI single driver with No Cache, but I've tried a variety of other settings on the NVMe - SSD Emulation, Discard, IO Thread, Async IO. I also tried both 2 and 3 on Proxmox installed on both ext4 and xfs.
I'm new to Proxmox, but the contradiction between fio benchmarks and the performance of the MySql process just don't make sense to me.
Does anyone have any idea what sort of issue could cause this?
Thanks.
I've got a process that loads data into a mysql database - it's essentially several gigs worth of batched inserts. It runs pretty consistently in each environment.
These are the three environments I'm comparing:
1. MySql in a Docker container on Ubuntu, bare metal, with the Docker volume on a local SATA drive.
- This is the machine I've been running this process on for years.
- The process takes about an hour.
2. MySql in a Docker container on an Ubuntu Cloud Image Proxmox VM with the VM image on a Synology NFS, 1GbE.
- The NFS just has a single, old, SATA drive, because I wanted to test the network impact on the process.
- The process runs about 10% slower than the original setup, which seems like a reasonable trade-off to me, especially on 1GbE.
3. MySql on an Ubuntu Cloud Image Proxmox VM with the VM image on the machine's NVMe.
- I've tried this setup with MySql both directly on Ubuntu, and in a Docker container
- fio benchmarks about 10x better on this VM than the VM on NFS, which is understandable.
- The process takes about 100 minutes, which seems really strange, especially considering how well the NFS setup did.
Scenarios 2 and 3 are the same machine, different VMs.
I've only tried the VirtIO SCSI single driver with No Cache, but I've tried a variety of other settings on the NVMe - SSD Emulation, Discard, IO Thread, Async IO. I also tried both 2 and 3 on Proxmox installed on both ext4 and xfs.
I'm new to Proxmox, but the contradiction between fio benchmarks and the performance of the MySql process just don't make sense to me.
Does anyone have any idea what sort of issue could cause this?
Thanks.
Last edited: