Hi all.
This is probably more of a HA list, or possibly even linux-practices, question but all hosts concerned are running CentOS and I reckon some of you guys might have some good suggestions. Feel free to tell me to piss off. :)
I'm building a new CentOS, DRBD, Heartbeat and NFS HA cluster. We already have boxes running similar setups on FC2/3 running NFS v2/3 with the Ultra Monkey Heartbeat packages. These new systems are using CentOS packages as much as possible.
[root@tempmailstore1a ha.d]# rpm -qa | grep -Ei 'heartbeat|drbd' heartbeat-pils-1.2.3.cvs.20050927-1.centos4 heartbeat-ldirectord-1.2.3.cvs.20050927-1.centos4 heartbeat-stonith-1.2.3.cvs.20050927-1.centos4 heartbeat-1.2.3.cvs.20050927-1.centos4 kernel-module-drbd-2.6.9-22.0.1.ELsmp-0.7.14-1.centos4 drbd-0.7.14-1.centos4
I'm messing around with NFS v4 (any and all experience/horror stories about EL4's NFS v4 in production environments more than welcome as I'm still at the evaluation/test stage) and looking at the best way to setup the --bind mounts for the NFS v4 pseudofilesystem.
So far in /etc/ha.d/haresources I have...
tempmailstore1a IPaddr::192.168.24.73 \ drbddisk::r0 \ Filesystem::/dev/drbd0::/mnt/drbd::ext3::rw,acl,data=journal \ portmap \ nfslock \ nfs \ rpcidmapd
Now, after the Filesystem resource.d script's called ideally I'd like to setup the --bind mounts as detailed here...
http://www.citi.umich.edu/projects/nfsv4/linux/using-nfsv4.html
As follows...
mount --bind /mnt/drbd /export/drbd
I would've thought the best way to achieve this, since it's effectively another filesystem of type "none" when mounted would be with the resource.d/Filesystem script... which can't take a --mount option.
Obviously I can hack around in the shell script or write my own, has anyone else bumped up against this? And how did you deal with it.
Will.
On 08/02/06, Will McDonald wmcdonald@gmail.com wrote:
I'm messing around with NFS v4 (any and all experience/horror stories about EL4's NFS v4 in production environments more than welcome as I'm still at the evaluation/test stage) and looking at the best way to setup the --bind mounts for the NFS v4 pseudofilesystem.
Just found the answer to my own question...
https://www.redhat.com/archives/fedora-list/2003-October/msg00957.html
"mount --bind" is the same as "mount -o bind"
So I can just have bind as an option in haresources for that particular mount.
I'd still be interested in hearing other's opinions on NFS v4 in production environments.
Will.
On Wed, 2006-02-08 at 14:08 +0000, Will McDonald wrote:
On 08/02/06, Will McDonald wmcdonald@gmail.com wrote:
I'm messing around with NFS v4 (any and all experience/horror stories about EL4's NFS v4 in production environments more than welcome as I'm still at the evaluation/test stage) and looking at the best way to setup the --bind mounts for the NFS v4 pseudofilesystem.
Just found the answer to my own question...
https://www.redhat.com/archives/fedora-list/2003-October/msg00957.html
"mount --bind" is the same as "mount -o bind"
So I can just have bind as an option in haresources for that particular mount.
I was just writing that :)
I'd still be interested in hearing other's opinions on NFS v4 in production environments.
I have read about, but haven't had any major NFS v4 issues. I don't do a lot of NFS though ... in fact, the only time I use NFS is if I am rebuilding a huge number of RPMS and using our round robin build script so I can get more than 1 server involved in the building.
I did package and build all the DRBD / Heartbeat tools for CentOS and I have been using these tools in production with zero problems since about 3 months before they were released as extras.
On 08/02/06, Johnny Hughes mailing-lists@hughesjr.com wrote:
On Wed, 2006-02-08 at 14:08 +0000, Will McDonald wrote:
Just found the answer to my own question...
https://www.redhat.com/archives/fedora-list/2003-October/msg00957.html
"mount --bind" is the same as "mount -o bind"
So I can just have bind as an option in haresources for that particular mount.
I was just writing that :)
Heh. Thanks.
I'd still be interested in hearing other's opinions on NFS v4 in production environments.
I have read about, but haven't had any major NFS v4 issues. I don't do a lot of NFS though ... in fact, the only time I use NFS is if I am rebuilding a huge number of RPMS and using our round robin build script so I can get more than 1 server involved in the building.
We'll be replacing the existing FC boxes with these new CentOS boxes (to take advantage of the less frenetic release cycle) to act as central NFS mailstores for Qmail/VPOPMail maildirs. I'll be sure to give them a week or two's hammering before switching.
I did package and build all the DRBD / Heartbeat tools for CentOS and I have been using these tools in production with zero problems since about 3 months before they were released as extras.
Much appreciated it is too. I was trying to avoid veering from base upstream packages... until I saw those. And Postgres 8 in the dev repo. And all the useful Dag things. :)
Prompted by Dag's recent robbery, I've actually been trying to hit up management for a modest OSS development fund, even if it's only a few hundred quid a year, to donate to the projects we benefit from. We'll see how that pans out. :)
Will.
On 08/02/06, Will McDonald wmcdonald@gmail.com wrote:
"mount --bind" is the same as "mount -o bind"
So I can just have bind as an option in haresources for that particular mount.
I was just writing that :)
Heh. Thanks.
Damn. It appears the Heartbeat resource.d/Filesystem script's checking to ensure the first device is a valid block device. :o\
case $DEVICE in -*) # Oh... An option to mount instead... Typically -U or -L ;; [^/]*:/*) # An NFS filesystem specification... ;; *) if [ ! -b "$DEVICE" ] ; then ha_log "ERROR: Couldn't find device $DEVICE. Expected /dev/??? to exist" usage exit 1 fi;;
Ah well. I'll just make a Filesystem.local for --bind mounts and comment the block device check.
Will.
-
Johnny Hughes -
Will McDonald