<br><font size=2><tt>Lars,</tt></font>
<br>
<br><font size=2><tt>Thanks for the reply. See below.</tt></font>
<br>
<br><font size=2><tt>> On Tue, Feb 17, 2009 at 03:52:16PM -0700, David.Livingstone@cn.ca
wrote:<br>
> > Hello,<br>
> ><br>
> > I currently have two two-node clusters running heartbeat and
drbd(see<br>
> > background below). I also have a two-node test which I decided
to update<br>
> > to the latest releases of all. In so doing I downloaded and installed<br>
> > drbd 8.3.0(drbd-8.3.0.tar.gz) which includes three-node setups
using<br>
> > stacked clusters. Specifically havng a third backup/brp node<br>
> > geograghically removed from our production cluster is very appealing.<br>
> ><br>
> > I have looked at the online manual(http://www.drbd.org/users-guide/)<br>
> > and read the current information for three-node setups and have
some<br>
> > observations/questions :<br>
> > - An illustraion/figure of a three-node setup would help.<br>
</tt></font>
<br><font size=2><tt>> there are several ways to to it.<br>
> you can also have four nodes: two two-node DRBD, the primary of which
is<br>
> the "lower" resource of a "stacked" DRBD.</tt></font>
<br>
<br><font size=2><tt>Are there some examples I can review somewhere ?<br>
</tt></font>
<br><font size=2><tt>> > - From your "Creating a three-node
setup" example on which machine does<br>
> > the stacked-on-top-of address run(ie 192.168.42.1) ?</tt></font>
<br>
<br><font size=2><tt>> IP should be managed by heartbeat/pacemaker.
it needs to be present<br>
> before you promote the "upper" resource to Primary.<br>
</tt></font>
<br><font size=2><tt>> > In my case my third<br>
> > node is not on the same ip segment as my two other nodes.<br>
</tt></font>
<br><font size=2><tt>> no matter.</tt></font>
<br>
<br><font size=2><tt>So in the "Creating a three-node setup"
the ip would only ever be assigned</tt></font>
<br><font size=2><tt>to resource r0 nodes(alice or bob). Correct ?<br>
</tt></font>
<br><font size=2><tt>> > - After doing some searching I hit on the
http://drbd-plus.linbit.com<br>
> > page which mentions configuraion keywords "ignore-on"
and "use-csums".<br>
> > Neither of these exist in the drbd.conf man page. Are they needed
?<br>
</tt></font>
<br><font size=2><tt>> solved differently.<br>
> ignore-on was not flexible enough, so it was dropped.<br>
> use-csums has been replaced with csums-alg (so you can chose the<br>
> algorithm to be used for the checksum based resync).</tt></font>
<br>
<br><font size=2><tt>okay<br>
</tt></font>
<br><font size=2><tt>> > - The manual talks about the drbdupper resource
used in R1 style<br>
> > clusters. What about CRM style clusters ?<br>
</tt></font>
<br><font size=2><tt>> "interessting" setups with "interessting"
constraints.<br>
> or use drbdupper resource anyways.<br>
> We probably need a blog post or other feature about this.</tt></font>
<br>
<br><font size=2><tt>That would be great. Any examples ? </tt></font>
<br>
<br><font size=2><tt>> > - In the R1 style configuraion you state
:<br>
> > "The third node, which is set aside from the Heartbeat cluster,
will<br>
> > have the other half of the stacked resource available permanently."<br>
> > I presume by this you mean that if the two-node cluster disappears
that<br>
> > the mounting/application startup on the backup node is done manually
?<br>
</tt></font>
<br><font size=2><tt>> more or less, yes.<br>
</tt></font>
<br><font size=2><tt>> > Other Questions :<br>
> > - Is the manual available for download/printing ?<br>
</tt></font>
<br><font size=2><tt>> No. We hand it out in training sessions, though.</tt></font>
<br>
<br><font size=2><tt>Vienna sounds good ... now if I could convince my
boss ...<br>
</tt></font>
<br><font size=2><tt>> > - Has anyone used the nx_lsa(Linx Sockets
Acceration) driver to run drbd ?<br>
</tt></font>
<br><font size=2><tt>> I'm not exactly sure what that is supposed to
do.</tt></font>
<br>
<br><font size=2><tt>See http://www.netxen.com/technology/pdfs/Netxen_LinuxSocketsAcc_r3.pdf</tt></font>
<br><font size=2><tt>Essentially it implements a socket-level offload of
the network subsystem to</tt></font>
<br><font size=2><tt>a TCP stack running in firmware on the NIC. By using
the nxoffload facility</tt></font>
<br><font size=2><tt>you can specify tcp ip, ports or applications to offload.</tt></font>
<br><font size=2><tt><br>
</tt></font>
<br><font size=2><tt>> > Background :<br>
> > 1. Current two-node production clusters :<br>
> > - HW : - Proliant DL380G5<br>
> > - Crossover for drbd : HP NC510C(NetXen) 10GB using nx_nic<br>
> > - SW : - RHEL5 and kernel-PAE-2.6.18-92.1.10.el5<br>
> > - drbd : drbd-8.2.6-3, drbd-km-2.6.18_92.1.10.el5PAE-8.2.6-3,<br>
> > - heartbeat/pacemaker :<br>
> > heartbeat-2.99.0-3.1<br>
> > heartbeat-common-2.99.0-3.1<br>
> > heartbeat-resources-2.99.0-3.1<br>
> > pacemaker-heartbeat-0.6.6-17.2<br>
> > pacemaker-pygui-1.4-5.1<br>
> ><br>
> > 1. Test two-node cluster :<br>
> > - HW : - Proliant DL380G4<br>
> > - SW : - Latest RHEL5 and kernel-2.6.18-128.1.1.el5<br>
> > - drbd : drbd-8.3.0-3, drbd-km-2.6.18_128.1.1.el5-8.3.0-3<br>
> > - heartbeat/pacemaker :<br>
> > heartbeat-2.99.2-6.1.i386.rpm<br>
> > heartbeat-common-2.99.2-6.1.i386.rpm<br>
> > heartbeat-resources-2.99.2-6.1.i386.rpm<br>
> > pacemaker-1.0.1-3.1.i386.rpm<br>
> > pacemaker-pygui-1.4-11.9.i386.rpm<br>
> ><br>
</tt></font>
<br><font size=2><tt>> <br>
> --<br>
> : Lars Ellenberg<br>
</tt></font>