R. Steve McKown [Mon, 9 Dec 2013 16:07:49 +0000 (09:07 -0700)]
BackupPC_ovz: control of debug output
Write to the debug file only if the debug file is already present at
script start. This is a simple way to control whether debug output is
generated or not.
The Backup_PC_ovz script will run the command which is the text starting
with the token '$pingPath', substituting '$host' for its HN if '$host'
is found to be a VE.
R. Steve McKown [Tue, 26 Nov 2013 16:58:10 +0000 (09:58 -0700)]
BackupPC_ovz: implement a debug output file
Debug is stored in the file /tmp/BackupPC_ovz.debug. For now, selecting
debug on or off requires editing the script. A better solution would be
to pass something in on the command line.
R. Steve McKown [Thu, 7 Feb 2013 23:27:47 +0000 (16:27 -0700)]
Dynamically find SCSIDEV
Working from the expected device, BASEDEV, determine the SCSI host value
by interrogating the /sys filesystem. This prevents registration from
working, but now the device should automatically register when plugged
in (hot swap drive on a sil3132 chipset). This allows the device to be
properly unregistered regardless of the attached scsi host.
R. Steve McKown [Tue, 29 Jan 2013 03:57:00 +0000 (20:57 -0700)]
esata: new known state partitioned
Properly determines a registered state independently of a partitioned
drive. Before, the script simply looked for /dev/sdX2 to determine
registration state. This would be inaccurate when a drive with no
partition table was registered.
R. Steve McKown [Fri, 18 Jan 2013 01:18:01 +0000 (18:18 -0700)]
bpcdump: capture all log output
When bpcdump starts, use a temporary file in /tmp to capture log output
from the info() and error() functions. Then, which the correct home for
the info file is created, move the temporary file to that name and
continue along.
Prior to this commit, all of the output from info() and error() prior to
INFO being redefined were not also captured into the info file.
R. Steve McKown [Tue, 15 Jan 2013 06:38:50 +0000 (23:38 -0700)]
BackupPC_ovz: increase snapshot size
Mail server writes nearly 2GB during the backup window, so the previous
setting of 1GB for the snapshot partition is too small. Because the
vecommon partition holds pretty much all the CTs, then the cumulative
write requirements of al CT's on that node using vecommon must be taken
into account when properly sizing the snapshot volume size.
25GB seems like a reasonable number for now. That is at the moment
about 10% of the free space in the volume group.
R. Steve McKown [Fri, 12 Oct 2012 23:27:45 +0000 (17:27 -0600)]
bpcdump: allow removal of any VE backup
When looking to free space for the new backup, allow removal of older
backups for any VE, not just for the current VEID. Since the typical
use case now is to back up just a single BackupPC container, a change in
VEID is going to be due to a replacement BackupPC container, such as in
a major upgrade scenario.
If these scripts are in fact used in the future to back up multiple
BackupPC containers, additional work will be required on the script.
R. Steve McKown [Fri, 12 Oct 2012 23:24:30 +0000 (17:24 -0600)]
bpcdump: correct defect that removes all data
If when looking for space to reclaim for a new backup the ls that scans
for old backups finds nothing, the empty result ends up causing a
recursive removal of the entire contents of the destination directory
(usually a mount point for a partition on a removeable drive).
The correct behavior is to error with a message indicating not enough
space in such situations.
Adding BackupPC_verifyPool to the backuppc scripts. This was written by
another user but is not included in the standard bpc distro. It seems
quite useful.
smckown [Fri, 11 Jan 2008 00:46:58 +0000 (00:46 +0000)]
Code has now been tested for restore operation. A VE can restore to itself,
another VE or a non-VE server. Also, remove the unneeded runServer_test sub.