Re: ceph OSD down+out =>health ok => remove =>PGsbackfilling... ?

Burkhard Linke <Burkhard.Linke@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> · Tue, 26 Apr 2016 12:38:26 +0200



    Hi,

    
    On 04/26/2016 12:32 PM, SCHAER Frederic
      wrote:

    
        Hi,
         
        One simple/quick
            question.
        In my ceph cluster, I
            had a disk wich was in predicted failure. It was so much in
            predicted failure that the ceph OSD daemon crashed.
         
        After the OSD crashed,
            ceph moved data correctly (or at least that’s what I
            thought), and a ceph –s was giving a “HEALTH_OK”.
        Perfect.
        I tride to tell ceph to
            mark the OSD down : it told me the OSD was already down…
            fine.
         
        Then I ran this :
        ID=43 ; ceph osd down $ID ; ceph auth del
            osd.$ID ; ceph osd rm $ID ; ceph osd crush remove osd.$ID
      
    
    *snipsnap*

    
    Removing the dead OSD entry changed the CRUSH weight for the host,
    resulting in a redistribution of the data.

    
    The easiest way to prevent this is setting the CRUSH weight to 0.0
    before removing the OSD. This will trigger backfilling only once.

    
    Regards,

    Burkhard.
  

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