DragonFly BSD
DragonFly users List (threaded) for 2010-04
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: Snapshots ordering on slave and pfs according to freequency for snapshot management

From: Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>
Date: Wed, 28 Apr 2010 11:23:25 -0700 (PDT)

:> Yeah, I'm in full agreement with you.
:> I think I was a bit misanderstood: Hammer performance is *much better* than
:> rsnapshot (obviously) but there is no way to easily tell it to keep its
:> snapshot distribution in different intervals for archiving purposes.
:> This is the one thing rsnapshot does extremely well; I was just trying to say
:> this is a good idea we should steal :-)
:hope Matt will agree ;-)

    A more complex snapshot retention scheme would be a nice project for
    someone.  The hammer config file is just a block of text so it is
    possible to add pretty much whatever we want to it, with accompanying

    If someone wants to take on this project I think it could be integrated
    into the hammer cleanup code fairly easily by having the hammer cleanup
    field classify the snapshots it makes itself using the comment field,
    and then parsing that field to figure out cleanup intervals.

    This is the kind of format I would recommend for the config entry:

    snapshots	5m 5d
    snapshots	1d 60d
    snapshots	1w 2y

    e.g. the above would mean:   Keep 5-minute snapshots for 5 days,
    daily snapshots for 60 days, and weekly snapshots for two years.
    (NOTE: This might not actually be viable, it is just an example,
    as you would have to have quite a bit of storage to keep weekly's
    for 2-years).

    hammer could track the snapshots independently, even if they overlap,
    by using the comment field to distinguish them.  It could then use the
    comment field when doing the prune to figure out which snapshots line
    is applicable.

					Matthew Dillon 

[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]