[DRBD-user] DRBD+ 3 node setup

[Dries Robberechts]

Hi,

I'm trying to go from a DRBD v7 2-node cluster to a DRBD+ v8 3-node cluster:

    portalblogsvalnfs001 - NFS server 1 - part of drbd_c0d1p1
    portalblogsvalnfs002 - NFS server 2 - part of drbd_c0d1p1
    portalblogsvalnfs000 - backup server - on stack of drbd_c0d1p1, part 
of drbd_c0d1p1_backup

The idea is that the backup server gets synchronized with the NFS 
servers, but when the actual backup starts it gets detached, mounts the 
data partition and the data is being copied to the backup tapes. After 
that, the data partition is unmounted and the drbd is attached again to 
the NFS servers for synchronization. This to avoid having to backup the 
actual NFS servers, which would generate a lot of filesystem I/O on the 
active production system.

Going from the 2 NFS node v7 cluster towards a 2 NFS node v8 cluster was 
relatively easy.

Add the backup server portalblogsvalnfs000 does not seem that easy, though.

When adding the second line to haresources, the 2 NFS servers stay in 
Secondary/Secondary (ip address hidden).

    portalblogsvalnfs001  drbddisk::drbd_c0d1p1 \
                          drbdupper::drbd_c0d1p1_backup \
                         
    Filesystem::/dev/drbd0::/export::ext3::noatime,nodiratime \
                          killnfsd \
                          nfs-common \
                          nfs-kernel-server \
                          Delay::3::0 \
                          IPaddr::IPADDRESS_PUBVIP/26/eth0

Also, starting drbd on the backup server (no NFS) ends up with quite a 
lot of errors:

    portalblogsvalnfs000:/root# /etc/init.d/drbd start
    /etc/drbd.conf:24: in resource drbd_c0d1p1, on portalblogsvalnfs001
    { ... } ... on portalblogsvalnfs002 { ... }:
            There are multiple host sections for the peer.
            Maybe misspelled local host name 'portalblogsvalnfs000'?
    /etc/drbd.conf:24: in resource drbd_c0d1p1, there is no host section
    for this host.
            Missing 'on portalblogsvalnfs000 {...}' ?
    /etc/drbd.conf:46: in resource drbd_c0d1p1_backup, my hostname
    (portalblogsvalnfs000) not found in referenced resource drbd_c0d1p1
    Starting DRBD resources:    /etc/drbd.conf:24: in resource
    drbd_c0d1p1, on portalblogsvalnfs001 { ... } ... on
    portalblogsvalnfs002 { ... }:
            There are multiple host sections for the peer.
            Maybe misspelled local host name 'portalblogsvalnfs000'?
    /etc/drbd.conf:24: in resource drbd_c0d1p1, there is no host section
    for this host.
            Missing 'on portalblogsvalnfs000 {...}' ?
    /etc/drbd.conf:46: in resource drbd_c0d1p1_backup, my hostname
    (portalblogsvalnfs000) not found in referenced resource drbd_c0d1p1
    Can not load the drbd module.
    portalblogsvalnfs000:/root#

The actual drbd.conf can be found below.

I have more or less followed the idea as it's published on 
http://drbd-plus.linbit.com/examples:3nodes-setup
Maybe I overlooked something, or am I doing something wrong. I cannot 
seem to find a lot more tutorials on a 3 node setup, though.


Thanks, regards,
Dries.

My drbd.conf file (ip addresses hidden):

    resource drbd_c0d1p1 {
      protocol C;
      #incon-degr-cmd "halt -f";
      disk {
        on-io-error detach;
      }
      net {
      }
      syncer {
        rate 10M;
        #group 1;
        al-extents 257;
      }
      on portalblogsvalnfs001 {
        device     /dev/drbd0;
        disk       /dev/cciss/c0d1p1;
        address    IPADDRESS_INT001:7789;
        meta-disk  internal;
      }
      on portalblogsvalnfs002 {
        device     /dev/drbd0;
        disk       /dev/cciss/c0d1p1;
        address    IPADDRESS_INT002:7789;
        meta-disk  internal;
      }
    }

    resource drbd_c0d1p1_backup {
      protocol A;
      startup {
        degr-wfc-timeout 120;
      }
      disk {
        on-io-error pass_on;
      }
      net {
      }
      syncer {
        rate 60M;
        al-extents 257;
      }
      stacked-on-top-of drbd_c0d1p1 {
        device     /dev/drbd1;
        address    IPADDRESS_PUBVIP:7789;
      }
      on portalblogsvalnfs000 {
        device     /dev/drbd1;
        disk       /dev/cciss/c0d1p1;
        address    IPADDRESS_PUBSRV000:7788;
        meta-disk  internal;
      }
    }