Re: ceph balancer: further optimizations?

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

 



On Mon, Aug 20, 2018 at 10:45 PM Stefan Priebe - Profihost AG
<s.priebe@xxxxxxxxxxxx> wrote:
>
>
> Am 20.08.2018 um 22:38 schrieb Dan van der Ster:
> > On Mon, Aug 20, 2018 at 10:19 PM Stefan Priebe - Profihost AG
> > <s.priebe@xxxxxxxxxxxx> wrote:
> >>
> >>
> >> Am 20.08.2018 um 21:52 schrieb Sage Weil:
> >>> On Mon, 20 Aug 2018, Stefan Priebe - Profihost AG wrote:
> >>>> Hello,
> >>>>
> >>>> since loic seems to have left ceph development and his wunderful crush
> >>>> optimization tool isn'T working anymore i'm trying to get a good
> >>>> distribution with the ceph balancer.
> >>>>
> >>>> Sadly it does not work as good as i want.
> >>>>
> >>>> # ceph osd df | sort -k8
> >>>>
> >>>> show 75 to 83% Usage which is 8% difference which is too much for me.
> >>>> I'm optimization by bytes.
> >>>>
> >>>> # ceph balancer eval
> >>>> current cluster score 0.005420 (lower is better)
> >>>>
> >>>> # ceph balancer eval $OPT_NAME
> >>>> plan spriebe_2018-08-20_19:36 final score 0.005456 (lower is better)
> >>>>
> >>>> I'm unable to optimize further ;-( Is there any chance to optimize
> >>>> further even in case of more rebelancing?
> >>>
> >>> The scoring that the balancer module is doing is currently a hybrid of pg
> >>> count, bytes, and object count.  Picking a single metric might help a bit
> >>> (as those 3 things are not always perfectly aligned).
> >>
> >> Hi,
> >>
> >> ok i found a bug in the balancer code which seems to be present in all
> >> releases.
> >>
> >>  861                     best_ws = next_ws
> >>  862                     best_ow = next_ow
> >>
> >>
> >> should be:
> >>
> >>  861                     best_ws = copy.deepcopy(next_ws)
> >>  862                     best_ow = copy.deepcopy(next_ow)
> >>
> >> otherwise it does not use the best but the last.
> >
> > Interesting... does that change improve things?
>
> It fixes the following (mgr debug output):
> 2018-08-20 22:33:46.078525 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001180, misplacing 0.000912
> 2018-08-20 22:33:46.078574 7f2fbc3b6700  0 mgr[balancer] Score got
> worse, taking another step
> 2018-08-20 22:33:46.078770 7f2fbc3b6700  0 mgr[balancer] Balancing root
> default (pools ['cephstor2']) by bytes
> 2018-08-20 22:33:46.156326 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001180, misplacing 0.000912
> 2018-08-20 22:33:46.156374 7f2fbc3b6700  0 mgr[balancer] Score got
> worse, taking another step
> 2018-08-20 22:33:46.156581 7f2fbc3b6700  0 mgr[balancer] Balancing root
> default (pools ['cephstor2']) by bytes
> 2018-08-20 22:33:46.233818 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001180, misplacing 0.000912
> 2018-08-20 22:33:46.233868 7f2fbc3b6700  0 mgr[balancer] Score got
> worse, taking another step
> 2018-08-20 22:33:46.234043 7f2fbc3b6700  0 mgr[balancer] Balancing root
> default (pools ['cephstor2']) by bytes
> 2018-08-20 22:33:46.313212 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001180, misplacing 0.000912
> 2018-08-20 22:33:46.313714 7f2fbc3b6700  0 mgr[balancer] Score got
> worse, trying smaller step 0.000244
> 2018-08-20 22:33:46.313887 7f2fbc3b6700  0 mgr[balancer] Balancing root
> default (pools ['cephstor2']) by bytes
> 2018-08-20 22:33:46.391586 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001152, misplacing 0.001141
> 2018-08-20 22:33:46.393374 7f2fbc3b6700  0 mgr[balancer] Balancing root
> default (pools ['cephstor2']) by bytes
> 2018-08-20 22:33:46.473956 7f2fbc3b6700  0 mgr[balancer] Step result
> score 0.001152 -> 0.001180, misplacing 0.000912
> 2018-08-20 22:33:46.474001 7f2fbc3b6700  0 mgr[balancer] Score got
> worse, taking another step
> 2018-08-20 22:33:46.474046 7f2fbc3b6700  0 mgr[balancer] Success, score
> 0.001155 -> 0.001152
>
> BUT:
> # ceph balancer eval myplan
> plan myplan final score 0.001180 (lower is better)
>
> So the final plan does NOT contain the expected optimization. The
> deepcopy fixes it.
>
> After:
> # ceph balancer eval myplan
> plan myplan final score 0.001152 (lower is better)
>

OK that looks like a bug. Did you create a tracker or PR?

-- Dan


> >
> > Also, if most of your data is in one pool you can try ceph balancer
> > eval <pool-name>
>
> Already tried this doesn't help much.
>
> Greets,
> Stefan
>
>
> > -- dan
> >
> >>
> >> I'm also using this one:
> >> https://github.com/ceph/ceph/pull/20665/commits/c161a74ad6cf006cd9b33b40fd7705b67c170615
> >>
> >> to optimize by bytes only.
> >>
> >> Greets,
> >> Stefan
_______________________________________________
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