At this point I have realized two things. 1) tar accepts a stream of filenames in, not a data stream. 2) backuppc expects specifically a tar stream, not just a file stream(with a list of files preceding the data). 

To that end I have thrown out the idea of using tar and I have resorted to writing my own perl script that uses Archive::Tar. So far testing is going well with dummy data but I dont have it fully working yet. 


On 25 April 2013 22:15, Adam Goryachev <mailinglists@websitemanagers.com.au> wrote:
On 26/04/13 14:39, Les Mikesell wrote:
> On Thu, Apr 25, 2013 at 6:09 PM, Lord Sporkton <lordsporkton@gmail.com> wrote:
>> I'm aware of zmanda and several other backup options however at this time
>> this is what we have and this is what we are trying to leverage. Perhaps it
>> will turn out that writing to a flat file is the only option. But the nature
>> of the backuppc commands leads me to believe there is some possibility to
>> link streams. mysqldump streams into a file and backuppc "appears" to stream
>> from the local file, across the tunnel. This leads me to believe(and i could
>> of course be wrong due to some caveat that Im simply not aware of yet) that
>> If i could redirect backuppc to stream from a stream rather than file, i
>> could get it to work. Tar is of course capable of accepting either stream or
>> file as input and mysqldump is capable of outputing to either stream or
>> file. I suppose I will just have to play around with it more maybe.
> I don't think you are going to find a way to get backuppc to collect
> the output stream directly.  However you could use disk space on
> another machine for the intermediate file copy, using either a pipe
> over ssh or an nfs mount  and let backuppc pick it up from there.
I think the issue is that backuppc expects tar to not just provide a
stream of data, but a list of filenames with the data for each file
included. If you pipe the data into tar, I'm not sure that tar will be
able to know what the filename is.

I was wondering if you could instruct backuppc to use a custom tar
command to backup a named pipe something like this:

mysqlbackup > somepipe

backuppc is instructed to backup somepipe using the tar protocol, and
tar is instructed to handle the file like a normal file instead of just
backing it up as a named pipe.

The theory being that backuppc will see it is backing up a file called
"somepipe", but in reality, the file contents never exist on disk.

The challenge here is that the backup script "mysqlbackup > somepipe"
will not actually complete (because it can't write to the pipe) until
the backup is started. Probably you need backuppc to run a pre-backup
script which will trigger the mysqlbackup to start running
(disconnected/background so that backuppc will continue), then backuppc
will read the backup contents, and finish. You would need one "share"
for each DB, so 50 DB's means 50 shares, or potentially, you might be
able to have a directory of pipes, and the backup script can start the
50 backups in parallel (each one backgrounded, but each one doesn't
really start until the previous one is finished). The challenge is to
make sure tar will quit reading from the pipe at the end of the backup
output, and also it won't start reading from the next pipe before the
backup data starts to be sent there (possibly, depending on when/how tar
decides it has finished reading the file/pipe)

IMHO, this *might* work, but could also be fairly fragile, and may have
many unintended side-effects.

Potentially, a second option would be to use mysql replication to keep a
current copy of all live databases on a 'backup' machine. Then you can
simply stop the mysql server, backup the raw DB files with backuppc, and
then re-start the mysql server. The mysql server will then catch up from
all the remote replication partners, and continue. This also gives you a
possible source of more up to date backup data if some (not all) problem
happens on the live DB server.

Regards,
Adam

--
Adam Goryachev
Website Managers
www.websitemanagers.com.au


------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr
_______________________________________________
BackupPC-users mailing list
BackupPC-users@lists.sourceforge.net
List:    https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki:    http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/