Web UI unresponsive and high RAM usage during disk formats and wipes

[jhl]

I'm swapping some drives so I've been trying to do 2 different operations and running into major slowdowns and instability with both of them, particularly when running these operations on multiple drives at once.

1. Running "sg_format -F /dev/sdX" to remove Data Integrity Feature from the drives I'm installing. While running this, I get repeated console errors that the SMART information on the drives can't be accessed, which I think is expected while they are formatting. What's not expected is, while running this on 3 drives the web UI became unresponsive. At least I could check progress with "sg_format -v /dev/sdX" in the console. I'm pretty sure I've done this on earlier versions of Truenas (possibly Core) in the past without that behavior.

2. Wipe drives with zeroes using the button in the GUI Disks menu. I am trying to do 6 at once. In a VM with 12GB of RAM the WebUI and console both became unresponsive. The last thing I was able to do was run htop and see "middlewared" was maxing out my CPU. I needed to use that server so I rebooted and moved the disks to a bare metal server with 32GB of RAM to wipe them. This is currently running as expected with stable web UI access, but the 32GB of RAM did max out almost instantly which seems strange with basically nothing going on except wiping disks.

Anybody experienced this? Is it a new bug in Cobia?

 

[NugentS]

Please see forum rules - Hardware?
You also mention a VM, what hypervisor and how are the disks seen by the VM (assuming thats what happenning)?

 

[jhl]

The VM that became unusable while trying to do these tasks was on Proxmox on a Cisco server, and the disks were accessed directly by the VM via a PCI passed through LSI 9207 HBA controller

The bare metal system, which still is working but is seeing high RAM usage and slow UI performance as it wipes the disks, is a Supermicro H10 motherboard with Xeon v3, 32GB ECC DDR3 and Dell H730 controller in HBA mode. I understand this controller doesn't have "real" IT mode firmware and is not fully supported, but this is the system that has worked better out of the 2 configurations.

 

[NugentS]

OK - that kills most of my ideas. I am currently badblocking 5 drives on a very underpowered 1.6GHz 4c/4t 16GB System. CPU useage is 0-10% and memory is 2.3GB. This system is on Core

Are you on Scale or Core?

 

[jhl]

I’m on Scale 23.10.1 - I was also surprised that writing zeroes (or random data) to drives would use any significant amount of RAM or CPU and I think I did it on Core no problem in the past. I also thought sg_format used internal SMART functions and shouldn’t really load the system, but I guess Truenas is having some sort of error because the disk is busy and it keeps trying to poll the SMART data