Re: [ceph-users] Questions about bluestore

[Jorge Pinilla López]

There are 2 configs to set the size of your DB and 
WalBluestore_block_db_sizeBluestore_block_wal_size
If you have an SSD you should give as much space as you can to the DB and don't 
care about the Wal (Wal would always be placed in the fastest device) 

I am not sure about hot moving the DB but as far as you have replicas you can 
always remake the osd.
If SSD breaks and it's not possible to recover the data all your OSDs will 
break but as far as you have a replicas in other node you can remake the OSDs 
and they will rebalance.That's why you shouldn't use consumers SSD for rocksDB.

 Mensaje original De: Mario Giammarco <mgiamma...@gmail.com> 
Fecha: 14/10/17  10:54 AM  (GMT+01:00) Para: ceph-users 
<ceph-users@lists.ceph.com> Asunto: Re: [ceph-users] Questions about bluestore 
Nobody can help me? 

Il ven 6 ott 2017, 07:31 Mario Giammarco <mgiamma...@gmail.com> ha scritto:
Hello,I am trying Ceph luminous with Bluestore.
I create an osd:
ceph-disk prepare --bluestore /dev/sdg  --block.db /dev/sdf

and I see that on ssd it creates a partition of only 1g for block.db
So:
ceph-disk prepare --bluestore /dev/sdg --block.wal /dev/sdf --block.db /dev/sdf 


and again it creates two partitions, 1g and 500mb
It seems to me that they are too small the ssd is underutilized (docs says you 
need a ssd greater than 1g to put a block.db on)
Other two questions:
- if I already have an osd bluestore can I move later the db on ssd?- docs says 
that I can add several block.db of different osds on one ssd. But what happens 
if ssd breaks?
Thanks,Mario

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

Re: [ceph-users] Questions about bluestore

[Mario Giammarco]

Nobody can help me?

Il ven 6 ott 2017, 07:31 Mario Giammarco  ha scritto:

> Hello,
> I am trying Ceph luminous with Bluestore.
>
> I create an osd:
>
> ceph-disk prepare --bluestore /dev/sdg  --block.db /dev/sdf
>
> and I see that on ssd it creates a partition of only 1g for block.db
>
> So:
>
> ceph-disk prepare --bluestore /dev/sdg --block.wal /dev/sdf --block.db
> /dev/sdf
>
> and again it creates two partitions, 1g and 500mb
>
> It seems to me that they are too small the ssd is underutilized (docs says
> you need a ssd greater than 1g to put a block.db on)
>
> Other two questions:
>
> - if I already have an osd bluestore can I move later the db on ssd?
> - docs says that I can add several block.db of different osds on one ssd.
> But what happens if ssd breaks?
>
> Thanks,
> Mario
>
___
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com