RE: [uml-devel] New stackable COW files

[Peter B. <pe...@ra...>]

>> This would allow me to have a backing file shared by multiple users. The
>> use of cow stacking would permit smaller backups (more like incremental
>> backups).
> I don't quite under stand how you are planning on using it, could you
elaborate?


Here is my usage scenario.

Several UMLs all using the same backing file.

The UMLs are backed up by taking a copy of their COW files.

Over time their COW files grow larger, and the backups take more space and
the backups take longer and put more load on the server.

At this point I could add a stackable COW file.  The new COW file
(presumably) starts off small.

Overtime (as the COW file grows), I can stack on more COW files.  However,
administratively this becomes more difficult.  There are more files to
manage.  More links in the chain that can break (I need all the stacked COW
files and the backing file for the backups to be correct).

Thus, I would like a tool that collapses (merges) a series of cow files.
e.g. cow3,cow2,cow1,backing => cow3 (reset back to 0), cow1 (result of a
cow3, cow2 merge), backing.

Re: the command line options you specified:
$ uml_moo -top 3 -bottom 1 -o cow_merged cow4

Perhaps something to do an 'in place merge' (like the current -d option).

e.g.
$ uml_moo -list cow4
4:cow4[V3] -> 3:cow3[V3] -> 2:cow2[V3] -> 1:cow1[V2] -> 0:backingfile

so "uml_moo -top 3 -bottom 1 -d -r cow4" would put the merged version of
cow3, cow2 and cow1 in cow1 (the -d option).  And cow4 would be changed to
link to the merged file (cow1) (the -r for remove merged cow files which
would eliminate the administrative hassle of having 'unnecessary' cow
files).  So you'd be left with:
2:cow4[V3] -> 1:cow1[V2] -> 0:backingfile

Cheers, Peter