Hi Sage,
Thank you for your answer.
So, there is no anticipated problem with how I did ?
Does the 'data' pool performance affects directly my filesystem
performance, even if there is no file on it ?
Do I need to have the same performance policy on 'data' pools than on
the other pools ?
Can I use the fact that my base data pool 'data' is different than
my /real/ data pools to improve filesystem performance (something like
putting the 'data' pool on SSDs) ?
Regards
--
Thomas Lemarchand
Cloud Solutions SAS - Responsable des systèmes d'information
On jeu., 2014-11-13 at 08:33 -0800, Sage Weil wrote:
> On Thu, 13 Nov 2014, Thomas Lemarchand wrote:
> > Hi Ceph users,
> >
> > I need to have different filesystem trees in different pools, mainly for
> > security reasons.
> >
> > So I have ceph users (cephx) with specific access on specific pools.
> >
> > I have one metadata pool ('metadata') and tree data pools ('data',
> > 'wimi-files, 'wimi-recette-files').
> >
> > I used file layouts ( http://ceph.com/docs/master/cephfs/file-layouts/ )
> > to associate directories with pools.
> >
> > My filesystem looks like that :
> > Path -> associated pool
> >
> > / -> data
> > /prod -> wimi-files
> > /prod/... -> wimi-files
> > /recette -> wimi-recette-files
> > /recette/... -> wimi-recette-files
> >
> > Is it the best way to achieve what I need, since it's not possible to
> > have multiple CephFS on a Ceph cluster ?
> >
> > I ask this because my 'rados df' seems strange to me :
> >
> > pool name category KB objects clones
> > degraded unfound rd rd KB wr wr
> > KB
> > data - 0 9045499 0
> > 0 0 434686 434686 9294004 0
> > metadata - 58591 52681 0
> > 0 0 2168219 2403048804 16461385 180433628
> > wimi-files - 9006435331 10169214 0
> > 0 0 296284 2747513 19225407 9064999231
> > wimi-recette-files - 1036224 309167
> > 0 0 0 345223 1401472 658388
> > 1170762
> > total used 27404544372 19576561
> > total avail 78033398196
> > total space 105437942568
> >
> > As you can see, there are 9045499 objects in 'data' pool, while there
> > are only two directories ( 'prod', 'recette' ), and not a single file in
> > this pool.
> >
> > Anyone know how this works ?
>
> The MDS puts backtrace objects in the base data pool in order to
> facilitate fsck and lookup by ino even when the data is stored elsewhere.
>
> Other strategies that don't do this are possible, but they're more
> complicated, and we opted to keep it as simple as possible for now.
>
> sage
>
> >
> > Thanks in advance !
> >
> > Regards
> > --
> > Thomas Lemarchand
> > Cloud Solutions SAS - Responsable des syst?mes d'information
> >
> >
> >
> >
> >
> > --
> > This message has been scanned for viruses and
> > dangerous content by MailScanner, and is
> > believed to be clean.
> >
> > _______________________________________________
> > ceph-users mailing list
> > ceph-users@xxxxxxxxxxxxxx
> > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
> >
>
> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
>
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
