Hi,
I have as a home server a single Supermicro (Mostly for ECC RAM) machine with both PVE and PBS installed (PBS not as a VM but directly on the host). Whenever I am verifying the PBS backups, the VMs themselves are slow, but I can't see why that would be the case given that :
- The VMs (only 2 of them right now) as running off ZFS mirrored SDD
- The PBS stores its backup on different ZFS mirrored spinning drives (2 x 8TB). Yes, verifying backups is expected to be slow, since the drives are likely to be reaching their max throughput. But that's fine with me (takes about 14 hours. I can live with this for now)
- One of my VMs is OpenMediaVault (again, running on the SSDs). Accessing those files via the network is sloooow when the verifying is going on, but as I stop the verifying everything becomes as expected again.
- I can see in the Proxmox dashboard IO delays of 10% when verifying is going on.
- All drives are SATA 2
What explains the VMs running on SSDs being slow when the backup HDD are being hammered? I understand there is some comparing of backup vs data going on during verifying, but the SSDs are comparatively being left alone (as my performance graph seem to confirm) while the HDD are huffing and puffing.
I also checked network performance (the PBS is defined in PVE as 127.0.0.1) but no interface seems to be used in any significant fashion.
To be it seems there should be no relationship between HDD being hammered and the VMs being slow. Since they are on the same machine,
Any clue to where I should start looking?
I have as a home server a single Supermicro (Mostly for ECC RAM) machine with both PVE and PBS installed (PBS not as a VM but directly on the host). Whenever I am verifying the PBS backups, the VMs themselves are slow, but I can't see why that would be the case given that :
- The VMs (only 2 of them right now) as running off ZFS mirrored SDD
- The PBS stores its backup on different ZFS mirrored spinning drives (2 x 8TB). Yes, verifying backups is expected to be slow, since the drives are likely to be reaching their max throughput. But that's fine with me (takes about 14 hours. I can live with this for now)
- One of my VMs is OpenMediaVault (again, running on the SSDs). Accessing those files via the network is sloooow when the verifying is going on, but as I stop the verifying everything becomes as expected again.
- I can see in the Proxmox dashboard IO delays of 10% when verifying is going on.
- All drives are SATA 2
What explains the VMs running on SSDs being slow when the backup HDD are being hammered? I understand there is some comparing of backup vs data going on during verifying, but the SSDs are comparatively being left alone (as my performance graph seem to confirm) while the HDD are huffing and puffing.
I also checked network performance (the PBS is defined in PVE as 127.0.0.1) but no interface seems to be used in any significant fashion.
To be it seems there should be no relationship between HDD being hammered and the VMs being slow. Since they are on the same machine,
Any clue to where I should start looking?
