Hi JP,
Thanks for replay. I really appreciate it. I've examined your piece of
config file and most of all I accepted and understood. But my main
previous question was not "how" but "why".
For example, why you choose these parameters:
after-sb-0pri discard-zero-changes;
after-sb-1pri discard-secondary;
after-sb-2pri disconnect;
As far as I see the most common configuration consist of two nodes. If
one of them Primary and other Secondary and you have split brain how
your system will react? Secondary node will try to implement changes
from Primary node (after-sb-0pri discard-zero-changes) and Primary node
at the same time will try to discard all changes which Secondary node
doesn't have (after-sb-1pri discard-secondary). Do I understand it
correct? If so, I can't imagine what we will have finally. I see this
settings in lots of examples in the Internet and my question is "why".
Why lots of people chose it? Apparently, my understanding of split
brain behavior is wrong. Could you explain me, please, how a system from
two nodes will self recovery with these settings?
Thank you in advance,
Ivan. (Sydney)
On 01/13/2011 05:44 AM, J. Ryan Earl wrote:
Reply inline:
On Wed, Jan 12, 2011 at 8:50 AM, Dan Barker <[email protected]
<mailto:[email protected]>> wrote:
As to the I/O handling with DEBD+GFS2+NFS Primary/Primary, I do
not know the
answer. Maybe someone else on the list can help.
Dan, in Atlanta (the Top Poster)
Oh we got a RFC-1855 heretic here ;-)
-----Original Message-----
From: [email protected]
<mailto:[email protected]>
[mailto:[email protected]
<mailto:[email protected]>] On Behalf Of Ivan
Pavlenko
Sent: Tuesday, January 11, 2011 4:29 PM
To: [email protected] <mailto:[email protected]>
Subject: [DRBD-user] DRBD settings difficulties with understanding
Hi ALL,
I'm a new to DRBD and have some difficulties with understanding of
drbd
settings. I've been reading manuals hard but still have lots of
questions.
1) Configuring I/O error handling strategies I see detach method
is highly
recommended but pass_on is still default method. Why? What differences
between these methods? Why we have unrecommended method as default
instead
of the first one? I have
DEBD+GFS2+NFS Primary/Primary solution. What should I use in my
case? Why?
First off, break this into 2 parts: DRBD+GFS2 and GFS2+NFS. I'll
answer the DRBD+GFS2 part. The GFS2+NFS part can be examined without
DRBD. I'd suggest looking at:
http://sources.redhat.com/cluster/doc/nfscookbook.pdf
common {
startup {
wfc-timeout 0 ; # Wait forever for initial connection
degr-wfc-timeout 60; # Wait only 60 seconds if this node was a
degraded cluster
}
net {
# these split-brain options automatically repair common cases with
GFS2
# and at a minimum ensure integrity
after-sb-0pri discard-zero-changes;
after-sb-1pri discard-secondary;
after-sb-2pri disconnect;
cram-hmac-alg sha512;
shared-secret "replace this secret";
}
handlers {
split-brain "/usr/lib/drbd/notify-split-brain.sh [email protected]";
}
disk {
# be safe, go into diskless mode
on-io-error detach;
}
syncer {
verify-alg md5;
}
}
In your resource you need this to make it dual-primary:
resource some-resource {
[...]
net {
allow-two-primaries;
}
startup {
become-primary-on both;
}
}
That what you need from DRBD for GFS2. You'll need to assure fencing
somehow. I personally use power-fencing. Once you have the DRBD+GFS2
specifics above down, you probably want to research GFS2+NFS.
-JR
_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user
