On Friday 16 December 2011 03:18:16 Allan Gottlieb wrote: > I need to add space to /var (thank you, libreoffice), which is on lvm. > Since my one volume group vg is getting low, I thought this would be a > good time to extend it as well.
Actually, you need space in /var/tmp/portage > Alan (McKinnon) has posted very helpful lvm bits (reprinted below). > Following alan's bottom up creation mandate I believe the idea is > > phy disk: /dev/sda my only drive > phy part: fdisk create another partition of type LVM (/dev/sda8) > phy vol: pvcreate /dev/sda8 > vol grp: vgextend vg /dev/sda8 > log vol: lvextend --size +10G /dev/vg/var > file sys: resize2fs /dev/vg/var > files/dirs: not relevant > > Questions > > 1. Apparently 2.6 (hence 3.x) kernels can expand mounted file systems > (/var is mounted as ext3). > > Since I can't unmount /var because it is in use, I guess that, if I > every need to shrink /var, I would need to boot off a CD. Is that > correct? For shrinking, you need to umount the partition first. Online-resizing only works when increasing the size. > Back in the day, we had single user mode for this, but I don't see > how to get the equivalent now. "rc single" (See Dales email) > Is it really safe to extend /var (i.e., /dev/vg/var) while mounted > as ext3? It sounds frightening since daemons could start running > and access /var. Yes, I regularly extend filesystems while already copying data to them. > 2. Since currently /var is entirely from /dev/sda7 (my original lvm > partition) should I use the optional parameter to lvextend to > force the new space for /var to come from there as well? > lv extend --size +10G /dev/sda7 No, LVM can assign space from any disk. It will know where to go when accessing it. If it's just temporary for LibreOffice, I'd suggest making a temporary LV, mount it at /var/tmp/portage and then emerging the software. Then when finished, umount and remove the LV. I see no need to have /var really large for normal use and have never tried shrinking filesystems myself. > thanks in advance Your welcome :) HTH, -- Joost
