[sheepdog] [PATCH] dog: check correct number of required nodes for newly created VDIs

[Hitoshi Mitake]

At Wed, 22 Jan 2014 09:30:47 +0900,
Hitoshi Mitake wrote:
> 
> At Tue, 21 Jan 2014 21:25:00 +0800,
> Liu Yuan wrote:
> > 
> > On Tue, Jan 21, 2014 at 09:43:11PM +0900, Hitoshi Mitake wrote:
> > > Current dog compares a number of copies of newly created VDIs and a
> > > number of nodes. But the number of copies should be compared to a
> > > number of zones when the VDIs are replicated disks.
> > > 
> > > Signed-off-by: Hitoshi Mitake <mitake.hitoshi at lab.ntt.co.jp>
> > > ---
> > >  dog/vdi.c |    7 +++++--
> > >  1 file changed, 5 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/dog/vdi.c b/dog/vdi.c
> > > index ec2c0b3..18d7f37 100644
> > > --- a/dog/vdi.c
> > > +++ b/dog/vdi.c
> > > @@ -580,6 +580,7 @@ static int vdi_create(int argc, char **argv)
> > >  	uint64_t oid;
> > >  	uint32_t idx, max_idx, ret, nr_copies = vdi_cmd_data.nr_copies;
> > >  	struct sd_inode *inode = NULL;
> > > +	int nr_required_nodes;
> > >  
> > >  	if (!argv[optind]) {
> > >  		sd_err("Please specify the VDI size");
> > > @@ -602,9 +603,11 @@ static int vdi_create(int argc, char **argv)
> > >  		return EXIT_USAGE;
> > >  	}
> > >  
> > > -	if (nr_copies > sd_nodes_nr) {
> > > +	nr_required_nodes = 0 < vdi_cmd_data.copy_policy ?
> > > +	                    sd_nodes_nr : sd_zones_nr;
> > 
> > why difference for ec and replication? zone concept is same for ec and
> > replication. you need need to check (nr_copies > sd_zones_nr) here, works
> > for ec and replication vdi.
> 
> Ah sorry for that. It is my silly mistake. I' ll send v2 later.

On the second thought, it seems taht we need no check on dog side. The
checking can be done in sheep side. In addition, sheep can check
better than dog because it knows the cluster is strict mode or not.

I'll remove the checking in the next version.

Thanks,
Hitoshi