[DRBD-user] rpc.lockd / rpc.statd on heartbeat takeover of drbd systems.

Todd Denniston Todd.Denniston at ssa.crane.navy.mil
Thu Mar 11 21:49:16 CET 2004


Please forgive the rambling disconnected nature of this message, I have spent
too many hours attempting to answer my own question. I yield, surely someone
on this list has figured out the answer to the question, but not got it in big
bold letters in the howto yet.

mainly thinking:
should I worry about getting lockd and or statd going correctly?
or should I just make sure the system is configured never to run statd?


what *all* does the script 'nfsserver' called out in the howto do? where can I
get a copy? It is not on my system, not described in the howto and not
apparent with a google search. Does/should it do anything with rpc.statd?
I have been running on the assumption that it is some other distribution's
version of fedora's /etc/rc.d/init.d/nfs script, and it seems to work in the
limited testing I have done.

on Fedora Core release 1, 
/etc/rc.d/init.d/nfs --- manages rpc.nfsd,  rpc.mountd & exportfs,
/etc/rc.d/init.d/nfslock --- manages rpc.lockd and rpc.statd

if I should worry about it, then should nfslock come before or after nfs in
the haresources line?



"Re: NFS SERVER TAKEOVER" [1] offers an incomplete analysis/recommendations.
"RE:  [NFS] Clustered NFS?" [2] looks like the drbd HOWTO gets many of the
preconditions ragnark mentions taken care of.

looks like starting rpc.statd with -n SHARED_NAME may be the way to go[3].
perhaps it should be the fully qualified DNS SHARED_NAME though[3].
and somewhere in this reading I found that putting 
'STATD_HOSTNAME=SHARED_NAME' in /etc/sysconfig/network should do the same
thing...after reading nfslock it looks like Fedora implements this.

Horms[4] indicates that starting rpc.statd the 'number of reboots' gets
incremented causing "the clients do what you want them to do - esabish new
file handles with the other nfs server."
Is this what we really want them to do? I thought we wanted the clients to
think nothing happened, though I am probably making the mistake of looking at
clients as the people at the keyboard.

[1]Thread starting with:
"Re: NFS SERVER TAKEOVER" Alan Robertson on 'Fri, 05 Oct 2001 21:23:07 -0600'
http://lists.community.tummy.com/pipermail/linux-ha-dev/2001-October/002627.html

[2]
"RE:  [NFS] Clustered NFS?" ragnark on '05/24/2002 18:28:39'
http://www.geocrawler.com/archives/3/789/2002/5/0/8754149/

[3]"Using the Linux NFS Client with Network Appliance Filers"
http://www.netapp.com/tech_library/3183.html?fmt=print
and specificaly section "5.3 Network Lock Manager"
http://www.netapp.com/tech_library/3183.html?fmt=print#5.3.

[4]"NFS as a heartbeat ressource (Was: NFS as a Cluster File System.)"
Horms on 'Sat, 8 Feb 2003 11:14:24 +0900"
http://lists.linux-ha.org/pipermail/linux-ha/2003-February/006480.html
sub thread started by Alan Robertson on 'Fri, 07 Feb 2003 08:03:34 -0700"
http://lists.linux-ha.org/pipermail/linux-ha/2003-February/006467.html


-- 
Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane) 
Harnessing the Power of Technology for the Warfighter



More information about the drbd-user mailing list