Thin pool does not start if VM has "-" in name

[daniel-ayers]

Qubes OS version:

4.0 (with current dom0 updates)

Affected component(s):

Storage (of AppVMs on Secondary Storage per instructions at https://www.qubes-os.org/doc/secondary-storage/)

---

Steps to reproduce the behavior:

1. Configure a new volume group and thin pool by following the instructions at https://www.qubes-os.org/doc/secondary-storage/.

   In my case I created a pool called ssdpool in a VG called ssd over two PVs, each of which were inside a LUKS container. Result:

[me@dom0 ~]$ sudo pvs
  PV                                                    VG         Fmt  Attr PSize   PFree 
  /dev/mapper/luks-01e0c688-dbc7-4366-858d-c29342b8f9f1 qubes_dom0 lvm2 a--  475.98g     0 
  /dev/mapper/sda1_crypt                                ssd        lvm2 a--  978.08g 46.34g
  /dev/mapper/sdb1_crypt                                ssd        lvm2 a--  931.51g     0 

[me@dom0 ~]$ sudo lvs
  LV                                      VG         Attr       LSize   Pool    Origin                                  Data%  Meta%  Move Log Cpy%Sync Convert
  root                                    qubes_dom0 -wi-ao---- 411.98g                                                                                        
  swap                                    qubes_dom0 -wi-ao----  64.00g                                                                                        
  ssdpool                                 ssd        twi-aotz--   1.82t                                                 0.12   0.50                            
  vm-another-test-private                 ssd        Vwi-a-tz--   2.00g ssdpool vm-another-test-private-1537743808-back 5.22                                   
  vm-another-test-private-1537743808-back ssd        Vwi-a-tz--   2.00g ssdpool                                         0.00                                   
  vm-another-test-private-snap            ssd        Vwi-aotz--   2.00g ssdpool vm-another-test-private                 5.27                                   
  vm-another-test-volatile                ssd        Vwi-aotz--  10.00g ssdpool                                         0.02                                   
  vm-baldrick-private                     ssd        Vwi-a-tz--   2.00g ssdpool                                         0.00                                   
  vm-file-library-private                 ssd        Vwi-a-tz--  97.66g ssdpool vm-file-library-private-1537743461-back 2.15                                   
  vm-file-library-private-1537743461-back ssd        Vwi-a-tz--  97.66g ssdpool                                         2.15  

2. Create a new VM using that pool, making sure the name of the VM includes a "-".

[me@dom0]$ qvm-create -P ssdpool --label red file-library

3. Reboot computer.

4. Attempt to start new VM - it will not start and message appears "Qube Status: file library Domain file-library has failed to start: volume ssd/vm-file-library-private missing"

5. Create another VM using the new thin pool, this time ensuring the name does not contain a "-".

[me@dom0]$ qvm-create -P ssdpool --label red baldrick

6. Reboot.

7. Attempt to start file-library - works.

Expected behavior:

Each VM starts as requested.

Actual behavior:

I tested this with various combinations of VMs using the new thin pool: only file-library (fails); file-library and baldrick (works); file-library and another-vm (fails); file-library, another-vm and baldrick (works).

If all of the VMs using the thin pool have "-" in the VM name the pool does not work/start.

General notes:

It appears there is a bug where the thin pool is not started unless there is at least one VM using it that does not have "-" in its name. Suggests a bug parsing VM or LV names for "-", noting that "-" is used as a delimiter in the LV names.

---

Related issues:

[esote]

Source for qvm-create is here and qvm-start here, if it helps with diagnosing your issue. Don't know if this is related to the command, or how Qubes starts the VMs.

[marmarek]

I've just tried and cannot reproduce the problem. Does it still happen on your system @daniel-ayers ?

[daniel-ayers]

I take it the commit means the problem was found? I'm happy to test it again with current released updates on my system if that helps.

[marmarek]

No, I haven't managed to reproduce it. Commit fixes similar issues in automated tests, but isolated to tests only (I've tried to use them too, to reproduce the problem).

[marmarek]

Maybe this isn't about '-' at all? Maybe it's because secondary VG + thin pool is activated after qubesd service start? That wouldn't explain why you had different results depending on VM name, but maybe that's just a coincidence?
What happens if, in the not working state, you restart qubesd service? Does it help?

[heinrich-ulbricht]

I've got comparable behavior and error message (Qubes 4.0, latest updates). VM on second HDD in thin pool does not start unless I start one VM on the boot HDD first. There is no dash in either of the both VM names.
@marmarek Restarting the qubesd service does indeed solve this! (sudo service qubesd restart in dom0)
Do you need any more input or is this sufficient for a fix? This is 100% reproducible.

[github-actions]

This issue is being closed because:

• This issue is on the "Release 4.0 updates" milestone.
• Qubes OS 4.0 reached EOL (end-of-life) over one year ago.
• There has not been any activity on this issue in over one year.

If anyone believes that this issue should be reopened and reassigned to an active milestone, please leave a brief comment.
(For example, if a bug still affects Qubes OS 4.1, then the comment "Affects 4.1" will suffice.)