On Thu, Sep 17, 2015 at 02:03:14PM +0200, Wolfgang Schweer wrote: > On Thu, Sep 17, 2015 at 11:54:21AM +0200, Giorgio Pioda wrote: > > On Thu, Sep 17, 2015 at 11:08:02AM +0200, Wolfgang Schweer wrote: > > > On Thu, Sep 17, 2015 at 08:30:37AM +0200, Petter Reinholdtsen wrote: > > > > > > > > Did the format of /proc/mounts change? > > > > > > The additional /dev/dm-* lines (added by the kernel?) might be causing > > > the reported failure. > > > > In WS client /proc/mounts the vg_system entries are totally mising. Only > > /dev/dm-X are there, which are not recognized by lvextend. > > Confirmed for a stock ws: /dev/dm-0 (for /) and /dev/dm-2 (for /usr) are > replacing the vg_system entries. > I guess the /proc/mounts change is due to initramfs-tools now doing a > file system check of the root fs (and of /usr if on a separate > partition). > > See /usr/share/doc/initramfs-tools/NEWS.Debian.gz This might work to solve the problem. Would someone with coding skills please check it.
--- a/debian-edu-fsautoresize 2015-09-20 14:11:49.000000000 +0200
+++ b/debian-edu-fsautoresize 2015-09-20 21:20:31.014572174 +0200
@@ -14,6 +14,7 @@
use Getopt::Std;
use Sys::Syslog qw(openlog syslog closelog LOG_NOTICE);
+use File::Basename;
# Using this module (instead of Filesys::DiskSpace) to get a version
# providing the device size, and not only free and used.
@@ -194,6 +195,12 @@
chomp;
my @f = split(/\s+/);
my $device = $f[0];
+ # Always use mapper names instead of kernel ones.
+ if (index ($f[0],"/dev/dm-") != -1) {
+ my $basename = basename($f[0]);
+ $device = "/dev/mapper/".`ls -l /dev/mapper/ | grep $basename | cut
-d' ' -f15`;
+ }
+ chomp $device;
my $mountpoint = $f[1];
my $typename = $f[2];
next unless (exists $fsops{$typename});
Wolfgang
signature.asc
Description: Digital signature
