On 7/18/2018 10:27 PM, Konstantin Shalygin wrote:
So mostly I want to confirm that is is safe to change the crush rule for
the EC pool.
Changing crush rules for replicated or ec pool is safe.
One thing is, when I was migrated from multiroot to device-classes I was
recreate ec pools and clone images with qemu-img for ec_overwrites
feature, so I don't have experience with changing erasure-profiles.
Ah right... in my case this is a rgw object store pool containing a few
PB data, so I don't want to try recreating it!
Your old profile doesn't have root value, so this is 'default', new rule
will be have 'default' too. So I don't see any disadvantages with this
migration.
There is a hidden change though, as the value of "default" changed
through time.
We originally had two crush root buckets, one "ssd" and the other
"default" (containing the hdd devices)... and no common root above these
to make device class rules work.
So I renamed "default" to "platter" (aka hdd); re-typed both that and
the "ssd" buckets arbitrarily to "rack", and moved them under a new
"default" root bucket. The old crush rules refer to the original ssd and
platter bucket ids, while the new device-class rules can refer to the
new default root.
Anyway I suspect the "crush-root=default" in the ec profile is probably
just used when creating the corresponding crush rule, to populate the
"take" command, so this shouldn't be significant.
At least I hope not. The relationship between the crush rule and the ec
profile, when changing the crush rule on an existing ec pool, is still
not very clear to me...
Graham
--
Graham Allan
Minnesota Supercomputing Institute - gta@xxxxxxx
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
