Does the Slave pfs take up more space that the Master?
Antonio Huete Jimenez
ahuete.devel at gmail.com
Fri Apr 16 03:30:39 PDT 2010
Hi Matt,
As 'hammer info' directive does, the only way I know to do this would
be to go through all the PFS until the max (HAMMER_MAX_PFS), if no one
has come with a better idea or solution.
Actually, this should be pretty straightforward if you just go through
all the PFS, identify the ones that exist and then you just simply
perform the cleanup on the @@ link. i.e:
antonioh at smash> hammer cleanup /@@-1:00007
cleanup /@@-1:00007 - handle PFS #7 using /var/hammer/@@-1:00007
snapshots - disabled
prune - run
rebalance - run..
reblock - run....
recopy - run....
Dunno tho if this solution would be optimal.
I would like to takeover this activity if no one is interesed.
Cheers,
Antonio Huete
2010/4/16 Matthew Dillon <dillon at apollo.backplane.com>:
> :What i remember happened to me was that master pfs are pruned/reblocked by
> :the periodic but the slave pfs are not, so time to time i used to get a lot
> :of used space in my backups, just go and prune/reblock by hand and check if
> :thats what happening to you, check the config before, so you sure not
> :deleting stuff you dont want to.
> :Damian
> :
> :On Mon, Apr 12, 2010 at 04:03, Siju George <sgeorge.ml at gmail.com> wrote:
> :
> :> Hi,
> :>
> :> I have two hammer file systems.
> :>
> :> Backup1 454G 281G 173G 62% /Backup1
> :> Backup2 454G 316G 138G 70% /Backup2
>
> Yes, I was trying to think of scenarios and I think you are close,
> that is likely the reason. The hammer cleanup code is only scanning
> mounted PFSs. I verified this by umounting my /var/crash (which is
> a PFS) and running hammer cleanup. It didn't check /var/crash's
> PFS.
>
> So its not master/slave, it is simply whether the PFS is mounted or
> not that is the problem.
>
> This is definitely a bug. hammer cleanup needs to go through all
> the PFSs. It would be a nice mini-programming-project for someone
> who wants to fix it in the hammer utility.
>
> -Matt
>
>
More information about the Users
mailing list