Bug #39307: EC pools with m=1 are created with an unsafe min_size by default - RADOS - Ceph

Actions

Copy link

Bug #39307

closed

EC pools with m=1 are created with an unsafe min_size by default

Added by Paul Emmerich about 5 years ago. Updated about 5 years ago.

Status:

Won't Fix

Priority:

High

Assignee:

Category:

Target version:

% Done:

Source:

Tags:

Backport:

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Component(RADOS):

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

Creating an EC pool with m=1 on 14.2.0 defaults to a min_size of k, e.g. min_size of 2 for a 2+1 pool.
Older versions defaulted to a min_size of 3.

I don't think we should go for a default that favors availability over durability ever. It should be a conscious choice by the user to run a cluster that accepts writes with no redundancy.

Actions

Copy link

Updated by Greg Farnum about 5 years ago

Priority changed from Normal to High

Actions

Copy link

Updated by Greg Farnum about 5 years ago

Hmm, is this a default EC mode or just something we let users set?

The change was deliberate in PR https://github.com/ceph/ceph/pull/26894 following discussion on https://github.com/ceph/ceph/pull/8008 and I'm not sure how much we want to prevent users shooting themselves in the foot.

Actions

Copy link

Updated by Sage Weil about 5 years ago

Status changed from New to Won't Fix

This was a deliberate choice. https://github.com/ceph/ceph/pull/26894 made teh change, based on a discussion on another PR, https://github.com/ceph/ceph/pull/8008

I think the udnerlying problem is that 2+1 is just a bad choice.. either you tolerate no failures at all or durability is also weak. I suggest we change the default EC profile to 2+2 instead?

Actions

Copy link