Re: Several questions about enbd

Peter T. Breuer <[email protected]> wrote:
> > Ok. What's the purpose of reassemble? Wouldn't just raidstart be enough?
> 
> I am afraid you are currently more familiar with the scripts than I am.
> I would have to go look, and it is late. I'll answer over the w/e.


I believe that this is what using mdadm instead of raidtls looks like
.... (untested).



diff -ur resource.d.org/enbd-primary resource.d/enbd-primary
--- resource.d.org/enbd-primary	Sun Feb 13 21:03:09 2005
+++ resource.d/enbd-primary	Mon Feb 14 00:37:17 2005
@@ -2,6 +2,7 @@
 # This file is "/etc/ha.d/resources.d/enbd-primary". It should be on
 # both secondary and primary in the failover pair.
 
+NBD=/dev/nd/a
 MD=/dev/md0
 MNT=/mnt
 MNTOPTIONS=""
@@ -39,14 +41,11 @@
            /etc/init.d/nbd restart client $NBDID
 
            sleep 2
-           mkraid --really-force --dangerous-no-resync $MD
-           raidsetfaulty $MD $PRIMARYPARTITION
+           mdadm -C -f --run $MD -l 1 -n 2 -x 0 $NBD/0 missing
            sleep 1
-           raidstop $MD
            sleep 1
-           raidstart $MD
            sleep 1
-           raidhotadd $MD $PRIMARYPARTITION
+           mdadm -a $MD $PRIMARYPARTITION
            mount -t $MNTTYPE $MD $MNT $MNTOPTIONS
         echo "."
 }
@@ -54,7 +58,7 @@
 primaryside_client_stop () {
         echo -n "Stopping enbd-primary and umounting"
             umount $MNT
-            raidstop $MD
+            mdadm -S $MD
 
             # kill primary
             /etc/init.d/nbd stop client $NBDID
diff -ur resource.d.org/enbd-secondary resource.d/enbd-secondary
--- resource.d.org/enbd-secondary	Sun Feb 13 21:22:01 2005
+++ resource.d/enbd-secondary	Sun Feb 13 21:28:18 2005
@@ -7,13 +7,16 @@
 
 MNT=/mnt
 MNTOPTIONS=""
-FSTYPE="ext3"
+MNTTYPE="ext3"
 
-PRIMARY="mahler"
 SECONDARY="strauss"
+SECONDARYPARTITION=/users/image
+
+PRIMARY="mahler"
+PRIMARYPARTITION=/dev/sda12
+
 
 NBDID="one"
-SECONDARYPARTITION=/users/image
 
 PATH=/sbin:/bin:/usr/sbin:/usr/bin;/usr/local/sbin:/usr/local/bin
 
@@ -25,7 +28,7 @@
 primaryside_server_start () {
         echo -n "Starting enbd-secondary takeover"
             # remove NBD component from raid mirror
-            raidsetfaulty $MD $NBD/0
+            mdadm -f $MD $NBD/0
             # shut down NBD primary daemon
             /etc/init.d/nbd stop client $NBDID
             # kill any lingering requests on the device
@@ -45,7 +49,7 @@
            fuser -muk $MNT
            umount $MNT
            # stop the whole raid device
-           raidstop $MD
+           mdadm -S $MD
            sleep 1
            # we need to make sure that the nbd-secondary is up
            # wait the 5s for the NBD device to reenable itself after
@@ -58,16 +63,13 @@
            # restart the raid device, but without resync. We will
            # then mark the nbd device fault and reintegrate it, 
            # causing a resync over the net
-           mkraid --really-force --dangerous-no-resync $MD
-           raidsetfaulty $MD $NBD/0
+           mdadm -C -f --run $MD -l 1 -n 2 -x 0 missing $PRIMARYPARTITION
            sleep 1
-           raidstop $MD
            sleep 1
-           raidstart $MD
            sleep 1
-           raidhotadd $MD $NBD/0
+           mdadm -a $MD $NBD/0
 
-           mount -t $FSTYPE $MD $MNT $MNTOPTIONS
+           mount -t $MNTTYPE $MD $MNT $MNTOPTIONS
         echo "."
 }
 

I just glanced through it to familiarize myself with it ... I don't see
why one has to run "create" (-C) all the time instead of "assemble"
(-A), but I can imagine that with the less precise raidtools only to
hand it was easier to do that way. With mdadm there should be
sufficient control to allow -A. Notice that I could get rid of many of
the commands.

In particular with mdadm one can start an array running in degraded
mode. I hope I did it right above.


I need to add a token to show which end had the server when both went
down :-). Isn't this handled by heartbeat, somehow? Surely it's
heartbeats's business! Can you look that up?

Peter