Backups
From CECWiki
Contents |
What gets backed up
- Home directories (includes H: network drive on Windows machines)
- Research directories
What doesn't get backed up
- /scratch and /tmp directories
- Local disks on Windows machines (i.e., C:)
Backup Levels
To understand how backups work, it helps greatly if you understand the idea of "backup levels". At the most basic level, backups can be classified as "Full" backups, or as "Incremental" backups.
Full Backups
"Full backups" are fairly self-explanatory: ALL the files on a disk are written to tape. Any file that existed at the time of the backup will be on tape and can be recovered, however since ALL files are in the archive, it may take several hours for staff to search the entire archive to find your particular file.
"Full" backups are also refered to as "Level 0" backups.
Incremental Backups
Incremental backups are commonly divided into nine levels, "Level 1" through "Level 9." (You do not need to know what level backup is involved for any given file, but understanding these "levels" is fundamental to understanding the system.)
When a level 1 backups is run, only files which have changed since the most recent full backup are archived. For example, if your home directory was fully backed up last weekend (a level 0 backup,) then tonight's level 1 backup of your home directory will only contain copies of files that you have modified in some way since that full backup was done last weekend.
Similarly, a level 2 backup will contain copies of any files which have changed since the most recent level 1 backup, and a level 3 backup will contain only copies of those files which were changed since the most recent level 2 backups, etc. etc.
File Recovery
There are two ways to get files/directories recovered; self recovery, or submitting a file recovery request.
Self recovery
This option is only available on filesystems mounted from DCS storage, and is not available on Windows machines.
The DCS filesystems have a hidden checkpoint (.ckpt/) directory which will not show up in any directory listing. Within the checkpoint directory, there should be three directories named for the date and time the particular checkpoint was created (i.e., 2008_02_16_12.06.03_EST). Within those directories are read only versions of the files as they existed at the time the checkpoint was made. You can just copy the files you want from the checkpoint directory into your normal directory. The checkpoint directory will start in whichever directory you enter it from.
Only three checkpoints are maintained, two 'daily' checkpoints and one 'weekly' checkpoint, so time is of the essence.
To tell if a directory is on a checkpointed filesystem you can open a terminal window and cd to the directory in question and do an ls .ckpt. If you are using a file browser, try adding .ckpt/ to the end of the location. It should give you a directory listing if the filesystem is being checkpointed, otherwise it will give you an error message. If you get an error message, then you'll need to send a file recovery request to cechelp@mtu.edu.
File recovery request
If it has been more than a week since the file existed (in the state you wanted it), or you only use Windows machines, or the file is on a filesystem that isn't getting checkpointed, then you need to sent a file recovery request to cechelp@mtu.edu.
In the recovery request you will need to include the following information:
- Your email address (unix login/email name)
- Full PATH filename of lost file or directory
- The computer the file was on
- Date and time the lost file was created
- Date and time the lost file was last modified
- Date and time the file was deleted
We need all this information so that we can find your file among the millions of files we back-up, or that get backed up by DCS. Here is how you can find out the hostname of the machine you are on:
- Under Windows:
Right click on "My Computer" and select "Properties".
Click on the "Computer Name" tab in the Property window that pops up.
Look for where it says "Full computer name:". To the right of that is the hostname. - Under Linux:
Right click on the desktop and select "Open Terminal Window".
In the terminal window, enter the command "hostname". - Under Solaris:
At the command prompt, enter the command "hostname".
If you're on the console of a Solaris machine, you know how to get a command prompt. - Under MacOS:
Very funny. We don't support MacOS.
For the full path filename:
- Under Windows:
It is what shows up in the "Address:" field at the top of the file browser window. - Under Linux:
It is what shows up in the "Location:" field at the top of the file browser window, or
if you are using a terminal window, you can enter the command "pwd" to find the directory you are in. - Under Solaris:
Enter the command "pwd" at the command prompt to find the directory you are in.
If you know the dates and times the file or directory was created, last modified, and deleted, that's great, write it down. If not, then the more you narrow it down, the easier it will be to find your lost file or directory. If it's only a matter of days between creation and deletion, dates and times can be crucial. If it's on the order of months or more between creation and modification, or between modification and deletion, you can just say "Unknown. Over X months before modification," or deletion as the case may be. If the file hadn't been on the system for over a day, then it's not likely that the file was backed up.
