See also bug #105936, it seems sort of relevant, although in theory it should be obsolete. It suggests a workaround of doing sync after lvcreate.
I used the test script from there, and it triggers the bug, although not consistently.
# bash /tmp/snapshot-collision.sh myvg 1 10 12M 4M
Dropping disk caches ...
Creating test LVs ...
Logical volume "poc1" created
Creating snapshots ...
Logical volume "poc1-1" created
mount: you must specify the filesystem type
Logical volume "poc1-7" created
Logical volume "poc1-2" created
Logical volume "poc1-5" created
[...]
This only happened once, and most other times I only got it to fail with
device-mapper: reload ioctl failed: Cannot allocate memory
Failed to suspend origin poc1
mount: you must specify the filesystem type
This is running on a VM with 4 GiB RAM, 3.5 GiB free. Should be enough to avoid memory allocation errors.
See also bug #105936, it seems sort of relevant, although in theory it should be obsolete. It suggests a workaround of doing sync after lvcreate.
I used the test script from there, and it triggers the bug, although not consistently.
# bash /tmp/snapshot- collision. sh myvg 1 10 12M 4M
Dropping disk caches ...
Creating test LVs ...
Logical volume "poc1" created
Creating snapshots ...
Logical volume "poc1-1" created
mount: you must specify the filesystem type
Logical volume "poc1-7" created
Logical volume "poc1-2" created
Logical volume "poc1-5" created
[...]
This only happened once, and most other times I only got it to fail with
device-mapper: reload ioctl failed: Cannot allocate memory
Failed to suspend origin poc1
mount: you must specify the filesystem type
This is running on a VM with 4 GiB RAM, 3.5 GiB free. Should be enough to avoid memory allocation errors.