How to handle split-brain?
Hi,
my first setup is working so far (replication 2 nodes). I wanted to test split-brain scenario so i disabled network between the 2 nodes and wrote on both of them and enabled network again. The log is now saying
my first setup is working so far (replication 2 nodes). I wanted to test split-brain scenario so i disabled network between the 2 nodes and wrote on both of them and enabled network again. The log is now saying
[2012-02-02 17:58:58.812234] I [afr-self-heal-common.c:705:afr_mark_sources] 0-test2-replicate-0: split-brain possible, no source detected
[2012-02-02 17:58:58.812289] W [fuse-bridge.c:419:fuse_attr_cbk] 0-glusterfs-fuse: 201: LOOKUP() / => -1 (Input/output error)
and the mounted volume is not accessible on both. How can i make node2 accepting that node1 has the better data and it should sync from there?
Thx
Mike
Update:
ok, i found a possible solution in the geo-replication manual. A simple rsync to make both bricks equal and doing a find on the brick to start self-heal. Is this the right way?
and the mounted volume is not accessible on both. How can i make node2 accepting that node1 has the better data and it should sync from there?
Thx
Mike
Update:
ok, i found a possible solution in the geo-replication manual. A simple rsync to make both bricks equal and doing a find on the brick to start self-heal. Is this the right way?
sync -PavhS --xattrs --ignore-existing /good-local-dir/ bad-node:/faulty-dir 
