Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Dear DRBD users,
I'm testing DRBD 9 always using the latest source code in order
test the environment and put it in production as soon as the code is
stable enough.
My initial problem is: my scenario demands at least 3 nodes and
2 primaries nodes, using a Distributed File system (I have choose
ocfs2). I have already configured the nodes and the sincronization and
the file system locking is working fine, but I couldn't find on DRBD 9
and 8.4 manuals the procedure to safely shutdown a primary node (without
produce a split-brain behavior) and to shutdown all nodes without furter
problems like nodes with standalone connections, outdate or inconsistent
disk states.
For example: when I want to shutdown all nodes I do the
following steps:
1. umount file system in all nodes
(umount /<mount-point>
2. shutdown the ocfs cluster
(/etc/init.d/o2cb unload)
3. shutdown the drbd cluster (drbadm
down <resource>)
But when I restart I get this situation (1 node
primary,uptodate and 2 nodes inconsistent,standalone) and the procedure
"--discard-my-data" as stated on manuals doesn't work as supposed to and
the only way to get all nodes working is recreating the metadata for
secondary nodes.
So bottom line: How to shutdown 1,2 or all nodes in a
drbd9/primary-primary/3-nodes scenario without any problems?
Thanks in advance,
Arthur Fernandes.