Sorry, I forgot to tell that those pgs assigned to the kill osd are still writable after I raise min_size from 1 to 2 but before I restarted the killed osd.
-------- Forwarding messages --------
From: "xxhdx1985126" <xxhdx1985126@xxxxxxx>
Date: 2016-10-09 18:08:45
To: "ceph-users@xxxxxxxx" <ceph-users@xxxxxxxx>
Subject: PG go "incomplete" after setting min_size
From: "xxhdx1985126" <xxhdx1985126@xxxxxxx>
Date: 2016-10-09 18:08:45
To: "ceph-users@xxxxxxxx" <ceph-users@xxxxxxxx>
Subject: PG go "incomplete" after setting min_size
Hi, everyone.
I'm a newbie about ceph and trying to do some test to understand the behavior of ceph. The following situation really overwhelmed me:
I first killed a osd, which made the size of the acting set of some pg became 1. Then I set min_size from 1 to 2, after which I started the killed osd. Then there came the phenomenon that all the pg previous assigned to the killed osd goes "incomplete". My cluster contains only 2 hosts, each running 10 osds. And I made the configurations that replicas of pgs be assigned to both osds.
Is it supposed to be this way? What is the philosophy about this?
Thank you:-)
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
