XCP-NG - storage/disk display. What's actually happening? Is this a case for passthrough?

Computer Hardware & Server Infrastructure Builds
1

I’m running xcp-ng 8.0. VM VDs are on NFS (TrueNAS).

So I have a CentOS VM with /home (ext3) on a local SSD SR (using an entire 480GB SSD), but the rest on NFS SR. df in CentOS reads /home as 11G used (3%), 400G available. But, df in the host with the uuid of SR (as well as looking at the SR in XenOrchestra, and XenCenter) as 281G used (68%), 138G available.

Basically, the reason I installed this drive in the first place was because I needed the i/o for a particular VM, and files (lots and lots of them) were going to be written/deleted, written/deleted, etc… which I understand could be torturous to an NFS SR.

So since the files are being deleted by a cron job in CentOS, 1) why are the OS and Xen showing such different values? I expect it’s because the OS deleting the file does not zero out the blocks on the SSD, or even the VD.

  1. Can someone explain what happens when you delete a file in a linux OS?
  2. Why isn’t the VD being updated, or is it more proper to ask, should the VD be updated?
  3. If the OS shows 3% usage, and the VD shows 68%, assuming I keep deleting the files, and /home in the OS stays at about 3%, but the VD grows and grows, what wil the VD actually do when it gets to 100%?
  4. Is there a better way to utilize this SSD for this purpose? Is passthrough a good choice here? If so, will df in the OS properly display accurate numbers? (I realize XO and xcp-Center will not be accurate, and will show 100% usage [according to Tom’s video])

Thoughts?

Thanks,

Rich

2

If I understand what you are asking, I don’t think there is any way to reclaim the space once the VDI expands. It will not be able to expand beyond the limits defined for the storage of that VM.

3

There are two storage parts to this… think of it as an SR for the whole VM except for /home, and an SR for just /home.

/home is the issue. It is an SSD, and the SR was an ext3 fs created in XO, it’s mounted like anything else in /etc/fstab. Here is an example of the issue. Let’s say /home starts with 1GB of data. Every day, 10GB of data gets written to /home. Then at the end of the day, a cron job deletes those files (10GB). df -h inside the CentOS VM shows 1GB used, but the xcp host does not, nor does XO or xcp-ng Center. They show 268GB used.

So, are you saying, if I have an 400GB SR, and if every day I write 10GB, then every night I delete ALL those files, I still can only write data for 40 days, and my storage will be full, even though I’m deleting those files?

Thanks