I was trying to run the ceph osd crush reweight command to move data out of one
node that has hardware failures and I noticed that as I set the crush reweight
to 0, some nodes would reflect it when I do ceph osd tree and some wouldn't.
What is the proper way to run command access cluster, does one need to run same
command *ceph osd crush reweight* from all mon nodes and it would push it down
to all osd tree and update the crush, or is it also ok to run it once on an osd
node and it will copy it across the other nodes and update the crush map?
ceph-users mailing list