Project

General

Profile

Bug #37590

api.vgcreate uses a PE size of 1G

Added by Jan Fajerski almost 4 years ago. Updated about 2 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Target version:
-
% Done:

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

Using such a large extend size leads to potentially unused space of just under 1G per devices.
Why was such a large PE size chosen?

Introduced by https://github.com/ceph/ceph/commit/4a1b97efc87f3df15a39a76de074b4791f3528ca

History

#1 Updated by Lars Marowsky-Brée almost 4 years ago

I agree, a 1GiB size for extents doesn't make a lot of sense; that it makes reading LVM2 output easier to read when looking at extents is not really a good reason, and I'm not sure how it allows us to create larger LVs? (This limitations no longer applies for LVM2 format vs LVM1, no?)

#2 Updated by Alfredo Deza almost 4 years ago

Is the problem here that this is not explained in more detail? Or is there a different issue that needs to be solved?

#3 Updated by Martin Weiss over 3 years ago

Is there any news on this topic? Do we really plan to waste so much of available disk capacity on the LVM based setups?

#4 Updated by Alfredo Deza over 3 years ago

There isn't anything different, we believe that the choice of 1GB for LVM extents made it an order of magnitude easier to calculate space available and the possible waste isn't significant enough to change back.

It becomes more of a problem on very small OSDs, but we aren't optimizing for that use case, so I would recommend pre-creating the LVs if using tiny OSDs is what is needed.

#5 Updated by Jan Fajerski about 2 years ago

  • Status changed from New to Fix Under Review
  • Pull request ID set to 34740

#6 Updated by Jan Fajerski about 2 years ago

  • Status changed from Fix Under Review to Resolved

Also available in: Atom PDF