Difference between revisions of "BackupPC Notes"
Line 11: | Line 11: | ||
## net - openssl | ## net - openssl | ||
## utils - bzip2 | ## utils - bzip2 | ||
− | |||
==Config Files== | ==Config Files== |
Revision as of 23:40, 7 October 2009
BackupPC Docs
Web Interface
The webpage that hold status messages for BackupPC is http://192.168.40.196/backuppc/ The address will change as the system goes live.
Adding a client for rsync
- Install cygwin w/ rsync on the client with the following packages
- admin - cygrunsrv
- net - rsync
- net - openssl
- utils - bzip2
Config Files
Most of the configuration for BackupPC is in /etc/backuppc/config.pl Many of the options can be over-ridden with host-specific files, but the config.pl should be edited to give a good default for Windows machines. Most importantly $Conf{BackupFilesExclude} or $Conf{RsyncArgs} need to be configured to exclude the followng items:
- Windows clients
- hiberfil.sys holds RAM contents during hibernation
- Temporary Internet Files IE cache
- Firefox Internet Cache
- RECYCLER Recyle Bin
- pagefile.sys Swap file
- System Volume Information System Restore files
- NTUSER.DAT Can cause problems with being in use??
- TEMP
- Linux clients
- /temp
- /dev
- /proc
- /sys
- /mnt
- /media Ubuntu mounts for removable media
- Firefox internet cache
- Other caches??
- Parts of /var??
Backup Method
Unless testing shows otherwise, the preferred method is rsync-over-ssh. This requires a minimal ssh and rsync install on the clients (probably a stripped-down cygwin). It has the advantage of being compressible and fully encrypted over the wire, and only transferring the changed parts of files. The downside is a higher CPU load on the server and clients for encryption and compression. For Linux hosts we should use a backuppc user and sudo privileges to avoid possible root-level compromises.
Storage Setup on Vienna
Vienna has 8x500GB drives set up with an XFS filesystem on LVM over Linux md RAID6. The device is mounted to /data and has 2.8TB of usable space. The RAID6 configuration allows up to 2 drives to fail without loss of data, and the LVM allows the filesystem to be expanded as new drives are added in the future.
Things we need to think about
- ssh authentication method (authorized_keys, host-based authentication)
- Open files (Outlook)
- Thousands of small files can take very long to index (Pulse Recording can accumulate >60k)
- Finding dhcp hosts reliably
- checksum seeding
- Parity Archive
- Database dumps from MySQL on webserver
- User notification policy??