specific OSD hostnames

["Lluís Pàmies i Juárez" <llpamies@xxxxxxxxxx>]

Hello,

I want to bind an OSD to a virtual network interface (let's call it
eth0:0). The setting is the following. There are two net interfaces:
 eth0   --> 10.0.0.100
 eth0:0 --> 10.0.0.200
"/etc/hostname" contains "testnode" and in "/etc/hosts" I have:
10.0.0.100 testnode
10.0.0.200 testnode0

Both monitor and osd have the following rule in /etc/ceph/ceph.conf:

[osd.0]
    host = testnode0
    cluster addr = 10.0.0.200
    public addr = 10.0.0.200

Then, I manually initialize the OSD (osd.0 in this example), and after
executing:

$ ceph osd crush add-bucket testnode0 host
$ ceph osd crush move testnode0 root=default
$ ceph osd crush add osd.0 1.0 host=testnode0

I can see the following crush map:

# buckets
host testnode0 {
     id -2 # do not change unnecessarily
     # weight 1.000
     alg straw
     hash 0 # rjenkins1
     item osd.0 weight 1.000
}
root default {
     id -1 # do not change unnecessarily
     # weight 1.000
     alg straw
     hash 0 # rjenkins1
     item testnode0 weight 1.000
}

However, after starting the OSD (start ceph-osd id=0), the OSD is
reallocated, changing the crushmap to:

# buckets
host testnode0 {
     id -2 # do not change unnecessarily
     # weight 0.000
     alg straw
     hash 0 # rjenkins1
}
host testnode {
     id -3 # do not change unnecessarily
     # weight 1.000
     alg straw
     hash 0 # rjenkins1
     item osd.0 weight 1.000
}
root default {
     id -1 # do not change unnecessarily
     # weight 1.000
     alg straw
     hash 0 # rjenkins1
     item testnode0 weight 0.000
     item testnode weight 1.000
}

Is there any way to avoid that and get the new OSD running under the
name "testnode0" ?

Thanks,

-- 
Lluís Pàmies i Juárez
http://lluis.pamies.cat
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com