Re: Using straw2 crush also with Hammer

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

 



On 12/11/15 09:37, Gregory Farnum wrote:
> On Wednesday, November 11, 2015, Wido den Hollander <wido@xxxxxxxx
> <mailto:wido@xxxxxxxx>> wrote:
> 
>     On 11/10/2015 09:49 PM, Vickey Singh wrote:
>     > On Mon, Nov 9, 2015 at 8:16 PM, Wido den Hollander <wido@xxxxxxxx
>     <javascript:;>> wrote:
>     >
>     >> On 11/09/2015 05:27 PM, Vickey Singh wrote:
>     >>> Hello Ceph Geeks
>     >>>
>     >>> Need your comments with my understanding on straw2.
>     >>>
>     >>>    - Is Straw2 better than straw ?
>     >>
>     >> It is not persé better then straw(1).
>     >>
>     >> straw2 distributes data better when not all OSDs are equally
>     >> sized/weighted.
>     >>
>     >>>    - Is it straw2 recommended  for production usage ?
>     >>>
>     >>
>     >> I'm using it in multiple clusters and it works fine.
>     >>
>     >
>     > Yes i followed your update on twitter  :)
>     >
>     >
>     >>
>     >>> I have a production Ceph Firefly cluster , that i am going to
>     upgrade to
>     >>> Ceph hammer pretty soon. Should i use straw2 for all my ceph pools ?
>     >>>
>     >>
>     >> I would upgrade to Hammer first and make sure that ALL clients
>     are updated.
>     >>
>     >> In case you are using KVM/Qemu, you will have to stop those processes
>     >> first and start them again before they are using the new code.
>     >>
>     >
>     > Thanks a lot for this pointer, i didn't knew this. So restarting
>     KVM / QEMU
>     > process effect running VMs ? ( some downtime )
>     >
> 
>     Yes. You can also (live) migrate to another host since that will spawn
>     Qemu with fresh code on the other host.
> 
>     But you have to make sure all running/connected clients support straw2
>     before you enable straw2.
> 
> 
> I believe straw2 only requires monitor support -- unlike the tuna led
> involved in executing CRUSH, straw2 is just about how the OSD/bucket
> weights get converted into a sort of "internal" straw weight. That's
> done on the monitors and encoded into the maps.
> 
> Right?
> -Greg

I don't believe that's the case.  If you convert a CRUSH map to use
straw2 then any connected QEMU/librbd clients without straw2 support
will die with something like:

terminate called after throwing an instance of
'ceph::buffer::malformed_input'
  what():  buffer::malformed_input: unsupported bucket algorithm: 5

Where:

ceph/src/crush/crush.h:CRUSH_BUCKET_STRAW2 = 5,


-- 
David Clarke
Systems Architect
Catalyst IT

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

[Index of Archives]     [Information on CEPH]     [Linux Filesystem Development]     [Ceph Development]     [Ceph Large]     [Ceph Dev]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [xfs]


  Powered by Linux