Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
I have a BIG doubt in how to proceed in restore a DRBD to normal
operation after a failed HD on node 1 (the primary).
The 2 nodes are two identical RedHat ES3 with DRBD 0.7.25, only with one
drbd resource, named vm1 (configured with internal metadata, and, on
both server, the local backing storage is on /dev/sdc2)
The situation is that :
1) The node 1 have a problem on /dev/sdc (is a HW raid 5, on witch 2
hard-drive have a fault at 10 second distance!).
2) The drbd automatically detach the backing storage.
3) So the node 1 state become :
cs:DiskLessClient st:Primary/Secondary ld:Inconsistent
And on node 2 the state become :
cs:ServerForDLess st:Secondary/Primary ld:Consistent
The applications (a VMWare server virtual machines) on node1 still
continue to work as usual, with virtual disk file (.vmdk) file on
/dev/drbd0.
So, now, for return back to a "safe" situation, i need to reboot the
node 1 for replace HD and reconfigure the Raid, but with minimum
downtime possible and of course, without lose the the data that is still
present on node 2.
Witch is the right procedure? Is correct that :
1) Stop the VMWare machine on node 1
2) Stop the drbd on node 1 with : drbdadm disconnect
3) Stop the drbd on node 2 with : drbdadm down
4) Shutdown the node 1 & change the hd
5) After boot-up node 1, is correct to do, on node 2 a drbd up, and on
node 1 ONLY a dbbd connect, so i don't attach and empty /dev/sdc1 device
that can destory the data on node2?
In fact, is like to start node 1 in diskless node.
6) After drbd connect on node1, i can promote the node 1 to primary also
without attached the local backing storage? I don't risk to destroy the
data on node 2 when promoting node 1 to primary? with :
drbdsetup /dev/drbd0 primary --do-what-I-say
Or i risk to destroy something?
7) After that, for return to use the local backing
storage also on node 1, is correct to do, ON NODE 1:
drbdadm attach vm1
drbdadm invalidate vm1
?
On the last online manual, of version 0.8.x, is written to do, as first
step when replacing an hd with a new one, create the metadata on it with :
drbdadm create-md resource
But my version 0.7.25 don't have this option on drbdadm, for crating the
metadata, and, anyway, the first time i have used the drbd, i don't
remember to have issued a commend to create the metadata, it will be
created automatically i think.
8) During the synchronization process, i can mount and use /dev/drbd on
node1, right? :)
Many thanks in advance for any help!
Regards.
Franco