holland-discuss team mailing list archive

[Tim Soderstrom <tim@xxxxxxxxxxxxxxxxxxxxx>]

Changes applied. The debug problem was a PEBCAK issue :) Also added in a simple estimated size check (by summing up the raw files) and some path checking. I know mysqldump does some fancy things to find the mysqldump binary and I was not able to figure that out. So instead I followed suit with the sqlite plugin and added a variable for telling Holland where the slapcat executable is.

On Jan 12, 2011, at 2:51 PM, Andrew Garner wrote:

> I'm not too familiar with ldap, but I know some people will backup the
> bdb backend directly as an alternative to slapcat.  You may be able to
> leverage LVM for that (although I suspect that may not be an option in
> your environment), and I believe there is a hotbackup utility for bdb
> as well that could be used.   Other backends have other implications
> of course (and IIRC, even slapcat won't work consistently for some
> backends on an active server).
> 
> A few minor things on your code:
> 
> * Stylistically, you probably want to use list.append, rather than
> list.insert(-1, ..) here
> * You don't need to cast to int() here: int(self.config['compression']['level'])
>   The configspec/validate_config() part takes care of that for you.
> That's one of the reasons holland has configspecs - so you don't have
> to convert strings all throughout your plugin.   I see this was taken
> from sqlite, but I thought I'd mention it.
> * You should close the output stream at the end.  It's actually
> forking a command (e.g. gzip) in the background, and close() will
> collect it and check the status.  Otherwise the plugin won't report
> errors from compression.  stream.close() can raise an IOError - this
> is does when the compression command exits non-zero.
> * You probably don't want to use os.setsid for subprocess here.
> * You probably don't want to actually run the slapcat command if
> dry_run is set (Logging list2cmdline may be useful though).
> 
> I'm not sure what you mean by DEBUG.  If your debug statements aren't
> showing up you probably need to run holland with debug logging:
> holland --log-level debug backup ... (or set level = debug in the
> holland.conf).   This should work fine - I just ran it yesterday
> during testing.
> 
> On Wed, Jan 12, 2011 at 1:34 PM, Tim Soderstrom
> <tim@xxxxxxxxxxxxxxxxxxxxx> wrote:
>> Howdy Folks!
>> 
>> After some Googling, I didn't see a really good backup tool for OpenLDAP. They probably exist and I just suck at Googling but it gave me a good chance to try using the Holland framework to do something both neat and actually useful for my job (which means I can work on it for my workey work time, yay!).
>> 
>> So I have created an OpenLDAP plugin that uses 'slapcat' to produce a backup. 'slapcat' is actually not very feature rich so I am wondering if other methods would be suitable (such as the OpenLDAP Python library) though documentation online seems to point to slapcat for a no frills get 'er down backup. In my testing I had to use it once or twice so it seems to do the trick.
>> 
>> Anyways, I have this in my Holland fork (https://github.com/m00dawg/holland). Any suggestions would be appreciated. My Python foo sort of went away pretty quickly after not using it for a few months so I ended up using the random and sqlite plugins as templates.
>> 
>> If you look at the source files, there not a lot of log debugging going on. I could not get log.DEBUG = ... working for some reason?
>> 
>> By the way, that virtual_env.py script freaking rocks!
>> _______________________________________________
>> Mailing list: https://launchpad.net/~holland-discuss
>> Post to     : holland-discuss@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~holland-discuss
>> More help   : https://help.launchpad.net/ListHelp
>> 
>