lvm_meta_size too small

[Michał Górniak]

When fuel creates lvm he assume that lvm_meta_size will be 64M and when i install on FC storage backend those size need to be bigger (example 128M).

If we wont change it deploy will always fail in my case 4mb was missing (we set 64M but metadata was 68M)

[Michael Polenchuk (mpolenchuk)]

"lvm_meta_size" is generated by fuel-agent (drivers/nailgun.py)

[Michał Górniak (p4cket)]

i found one more think.
when you run:
pvcreate --metacopiues 2 --metadatasize 64m /dev/sda4

then metadata size is 67120640
and it always this size when you use parameter metadatasize largen then 47m

[Alexander Gordeev (a-gordeev)]

> "lvm_meta_size" is generated by fuel-agent (drivers/nailgun.py)

That's not true.

lvm_meta_size is generated (well actually hardcoded to 64) by volume manager in nailgun https://github.com/openstack/fuel-web/blob/9bd38193ecf468ddb80e73d5ed40729b7216217d/nailgun/nailgun/extensions/volume_manager/manager.py#L745-L746

Then it gets passed to fuel-agent. And it turns out that fuel-agent is slightly adjusting it

https://github.com/openstack/fuel-agent/blob/master/fuel_agent/drivers/nailgun.py#L375-L397

for default lvm_meta_size=64, after the *precise* calculation it will be 28. That applies to all LVM volumes to be created during provisioning.

[Alexander Gordeev (a-gordeev)]

although i'm not sure, if a lvm volume could be created later during deployment.

Please ask fuel-library folks to confirm that bug is not related to fuel-library.

[Dmitry Pyzhov (dpyzhov)]

we don't backport medium bugs

[Bug Checker Bot (bug-checker)]

(This check performed automatically)
Please, make sure that bug description contains the following sections filled in with the appropriate data related to the bug you are describing:

actual result

version

expected result

steps to reproduce

For more detailed information on the contents of each of the listed sections see https://wiki.openstack.org/wiki/Fuel/How_to_contribute#Here_is_how_you_file_a_bug

[Alexander Gordeev (a-gordeev)]

I'm pretty sure that this bug actually duplicates https://bugs.launchpad.net/fuel/+bug/1546049

The issue is still the same, nailgun doesn't take into account that size of logical volumes always gets rounded to the nearest multiple of LVM extend size (which is 4MiB by default).

https://github.com/openstack/fuel-agent/blob/dcdd64a95245cdde57f1bd1e0a83720e6bf1f56a/fuel_agent/drivers/nailgun.py#L446-L458

But when, we're dealing with specific h/w/ such as FC or other tricky types of disk storages which expose relatively big optimal/minimal IO sizes, we might end up with the lack of free space in the middle of logical volumes creating due to volume boundaries alignment which will be done by LVM itself. Please consider https://bugs.launchpad.net/fuel/+bug/1546049/comments/4

Bearing in mind, the description of bug is very apt and there's no logs at all, proper RCA hasn't been done, i think it's safer to mark it as a duplicate.