Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I have a problem with a HA Cluster running on CentOS 5.4 x86_64.

Software used:

  • DRBD 8 (kmod-drbd-xen-8.0.16-5.el5_3.x86_64 / drbd-8.0.16-5.el5.centos)
  • Heartbeat
  • LVM2
  • Xen 3.1.2

My DRBD configuration syncs a logical volume. When I first create the meta data for DRBD on the logical volume, it starts syncing and works fine. The problem I am having is that it will not update the data after the first synchronisation, so basically you get a one time copy and than it stops.

When I trash the logical volume on either, re-create it and initialize the meta data for DRBD, it just starts syncing again (but again only once).

It gives no out of sync errors or anything, cat /proc/drbd shows me 2 healthy UpToDate nodes.

I've tried syncing the physical volume instead, this resulted in Xen not being able to boot any VM (mounted in the privileged domain).

It seems that it's not giving me any incremental updates but there are no errors what so ever.

Edit://

I've tried both a active/passive and a active/active setup.

I know it's not being synchronised by sample files being added to the ext3 filesystem running on the LVM that is configured with drbd.

Thanks for your help so far, attached my drbd config;

global { usage-count no; }
resource repdata {
  protocol C;
  startup { wfc-timeout 0; degr-wfc-timeout     120; }
  disk { on-io-error detach; } # or panic, ...
  net {
        allow-two-primaries;
        after-sb-0pri discard-younger-primary;
        after-sb-1pri consensus;
        after-sb-2pri call-pri-lost-after-sb;
        cram-hmac-alg "sha1";
        shared-secret "XXX"; }
  handlers {
        pri-on-incon-degr "echo o > /proc/sysrq-trigger ; halt -f";
        pri-lost-after-sb "echo o > /proc/sysrq-trigger ; halt -f";
  }
  syncer { rate 100M; }
  on node1.xxxxx.com {
    device /dev/drbd0;
    disk /dev/xenVG/vlayer;
    address 10.0.67.50:7788;
    meta-disk internal;
  }
  on node2.xxxxx.com {
    device /dev/drbd0;
    disk /dev/xenVG/vlayer;
    address 10.0.67.60:7788;
    meta-disk internal;
  }
}
share|improve this question
    
What leads you to believe that the data isn't being synchronised? –  womble Nov 10 '09 at 13:20
    
is it active/passive or active/active setup? also please paste drbd config –  Systemsninja Nov 10 '09 at 13:24
    
Where are your PVs located? –  Nils Mar 15 '12 at 20:37

1 Answer 1

I'm going to go out on a limb; from your description it sounds like you are mounting the volume on both DRBD participants.

You can't do this unless you are using a filesystem which supports shared-storage, such as OCFS2.

If you try do this with a filesystem like ext3 or XFS then you will discover huge inconsistencies in your data. Much like the behaviour which you describe.

share|improve this answer
    
Hi, this seems to indeed be the case. I tried to filter it in the lvm.conf and disable the VG on secondary node with "vgchange -a n", when I tried to start DRBD after this it couldn't find resources and therefore failed to start up. –  SBO Nov 10 '09 at 18:47
    
Then you'll either need to a) Change the filesystem, or b) Change your usage so that only one node has the block device(s) mounted at any one time. –  Dan Carley Nov 27 '09 at 10:52

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.