This is just a playground, right?
One approach would be to separate the small OSDs from the larger ones
by creating a new device class, e.g. "small" and leave "hdd" for the
larger drives. Then you can create a small pool for cephfs_metadata
(that usually doesn't require too much space) defined by a respective
crush rule pointing to the "small" devices. For the cephfs_data pool
you need more space so you use a crush rule that points to "hdd"
devices. This way you can decide for each pool (e.g. rbd) which OSDs
to use.
Zitat von Alokkumar Mahajan <alokkumar.mahajan(a)gmail.com>om>:
Hello Eugen,
Below is the o/p:-
ceph osd df:-
[image: image.png]
ceph osd tree:-
[image: image.png]
On Thu, 28 Nov 2019 at 14:54, Eugen Block <eblock(a)nde.ag> wrote:
> Hi,
>
> can you share the output of `ceph osd df` and `ceph osd tree`?
> The smallest of your OSDs will be the bottleneck. Since ceph tries to
> distribute the data evenly across all OSDs you won't be able to use
> the large OSDs, at least not without adjusting your setup.
>
> Regards,
> Eugen
>
>
> Zitat von Alokkumar Mahajan <alokkumar.mahajan(a)gmail.com>om>:
>
> > Thanks Wido.
> > @ But normally CephFS will be able to use all the space inside your
> > Cephcluster.
> > So, you are saying even if i see the size for CephFS pools as 55 GB, it
> can
> > still use whole 600GB (or the available disk) from Cluster?
> >
> > This is what i have with PGNum = 150 (for Data) and 32 (for Metadata) in
> my
> > cluster.
> >
> > Pool
> > Type
> > Size
> > Usage
> > cephfs_data
> > data
> > 55.5366GiB
> > 4%
> > cephfs_meta
> > metadata
> > 55.7469GiB
> >
> > Thanks
> >
> >
> > On Thu, 28 Nov 2019 at 13:49, Wido den Hollander <wido(a)42on.com> wrote:
> >
> >>
> >>
> >> On 11/28/19 6:41 AM, Alokkumar Mahajan wrote:
> >> > Hello,
> >> > I am new to Ceph and currently i am working on setting up CephFs and
> RBD
> >> > environment. I have successfully setup Ceph Cluster with 4 OSD's
(2
> >> > OSD's with size 50GB and 2 OSD's with size 300GB).
> >> >
> >> > But while setting up CephFs the size which i see allocated for CephFs
> >> > Data and metadata pools is 55GB. But i want to have 300GB assigned for
> >> > CephFs.
> >> >
> >> > I tried using "target_size_bytes" flag while creating pool
but it is
> >> > not working (it saus invalid command). Same result when i
> >> > use target_size_bytes with (ceph osd pool set) after creating pool.
> >> >
> >> > I am not sure if i am doing something silly here.
> >> >
> >> > Can someone please guide me on this?
> >> >
> >>
> >>
> >> You can set quotas on CephFS or on the RADOS pool for the CephFS
'data'
> >> (haven't tried the last one though).
> >>
> >> But normally CephFS will be able to use all the space inside your Ceph
> >> cluster.
> >>
> >> It's not that you can easily just allocate X GB/TB to CephFS.
> >>
> >> Wido
> >>
> >> > Thanks in adv.!
> >> >
> >> > _______________________________________________
> >> > ceph-users mailing list -- ceph-users(a)ceph.io
> >> > To unsubscribe send an email to ceph-users-leave(a)ceph.io
> >> >
> >> _______________________________________________
> >> ceph-users mailing list -- ceph-users(a)ceph.io
> >> To unsubscribe send an email to ceph-users-leave(a)ceph.io
> >>
>
>
> _______________________________________________
> ceph-users mailing list -- ceph-users(a)ceph.io
> To unsubscribe send an email to ceph-users-leave(a)ceph.io
>