From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bob Peterson Date: Thu, 6 Jun 2013 09:17:15 -0400 (EDT) Subject: [Cluster-devel] [PATCH 2/4] mkfs.gfs2: Align resource groups to RAID stripes In-Reply-To: <51B088C2.4080701@redhat.com> References: <1370520213-29676-1-git-send-email-anprice@redhat.com> <1370520213-29676-2-git-send-email-anprice@redhat.com> <83456593.48037538.1370523424334.JavaMail.root@redhat.com> <51B088C2.4080701@redhat.com> Message-ID: <1439287716.48101227.1370524635435.JavaMail.root@redhat.com> List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit | > In GFS1, we allowed mix-and-match resource group sizes, but we originally | > Do we still want to enforce this rule? | | Good question. I had assumed that we don't have a rule like that as the | rindex specifies the rg sizes. My next planned mkfs change is to allow | the journal creation code to ask for a resource group large enough to | contain all of a journal's data blocks so that they're always a single | extent. Returning to enforcing the rule would have implications for that | plan, too. | | Andy Hi, Some more thoughts to add to this discussion: First, since we're now trying to align our rgrps to raid stripes, maybe the rules no longer make sense. Second, we probably want to change gfs2_grow to ensure that new rgrps are also placed on the same stripe boundaries. If we enforce the uniform rgrp rule, this would probably happen automatically. If we don't, I bet gfs2_grow would need to change. Bob Peterson Red Hat File Systems