Mount options using ServiceGuard

[phil cook]

Hello - Can anybody suggest how I can control vxfs mount options for individual filesystems mounted by SG package scripts. A cluster I currently maintain mounts my Oracle data areas as so:

/dev/vg02/lvol1 /data/u01 vxfs log 0 0
/dev/vg02/lvol2 /data/u02 vxfs log 0 0

I would like to configure it so it mounts these datafile areas using raw i/o. i.e:

/dev/vg01/lvol8 /legend/oradata6 vxfs delaylog,mincache=direct,convosync=direct 0 0

There are other filesystems mounted by SG packages whose mount options I do not wish to affect.

Any help would be much appreciated.

Thanks.

[melvyn burnard · Accepted solution]

which version of Sg are you using?
the later versions have examples in the script files for mount options, as in:

# For example, if this package uses the file systems pkg1a and pkg1b,
# which are mounted on the logical volumes lvol1 and lvol2 with read and
# write options enter:
# LV[0]=/dev/vg01/lvol1; FS[0]=/pkg1a; FS_MOUNT_OPT[0]="-o rw"
# LV[1]=/dev/vg01/lvol2; FS[1]=/pkg1b; FS_MOUNT_OPT[1]="-o rw"

[phil cook]

Thanks for your reply. My pkg control scripts have this entry:

FILESYSTEMS
# Define the filesystems which are used by this package. The filesystems
# are defined as pairs of entries specifying the logical volume and the
# mount point for the file system. Each filesystem will be fsck'd prior
# to being mounted. The filesystems will be mounted in the order specified
# during package startup and will be unmounted in reverse order during
# package shutdown. Ensure that volume groups referenced by the logical
# volume definitions below are included in volume group definitions
# above.
#
# Example: LV[0]=/dev/vg01/lvol1; FS[0]=/pkg1
# LV[1]=/dev/pkg2/lvol1; FS[1]=/pkg2
#
LV[0]="/dev/vg02/lvol1"; FS[0]="/data/u01"
LV[1]="/dev/vg02/lvol2"; FS[1]="/data/u02"
LV[2]="/dev/vg03/lvol1"; FS[2]="/data/u03"
LV[3]="/dev/vg03/lvol4"; FS[3]="/data/u04"

So I suspect this means I am running an older version of SG? Bit disappointing.

Phil C

[Steve Lewis]

Phil,

You could always leave your filesystems out of that section altogether and mount them manually in the customer_defined_start/halt procedures. Just use the command you gave. I assume you have Online JFS (recent version and patches) required for the mincache=direct option.

Steve

[melvyn burnard]

so which SG version are you running?
it may be worthwhile you upgrading SG to the current Platform release of 11.09 for supportability, andthis will give you the new features.
to check, use swlist or do:
what /usr/lbin/cmcld

[phil cook]

Swlist reports the following:

lara:/usr/local/bin # swlist | grep Service
B3935BA A.10.10 MC / Service Guard
PHSS_15531 B.10.00.00.AA MC/ServiceGuard A.10.10 Cummulative patch
lara:/usr/local/bin #

So perhaps I should consider upgrading as you say.

Rgds

Phil

[melvyn burnard]

ServiceGuard 10.12 will ALSO give you the changes I mentioned, so not such a big upgrade path :-}
Also dont forget the latest SG patch for the veriosn you load, and remember, these patches are NEVER on the patch bundle CD's ;-}

HTH

[Carsten Krege]

The latest version for HPUX 10.20 is SG 10.12 (+patches). However, I'm pretty sure that the default 10.10 script provide the FS_MOUT_OPT arrays. You can check with

# cd /tmp
# cmmakepkg -s dummy
# vi dummy

You can then start to update your scripts.

In fact I would first check your present scripts and search in them for the string FS_MOUNT_OPT. If you see a line similar to

mount ${FS_MOUNT_OPT[$F]} $I ${FS[$F]}

in the function check_and_mount() the comments in the top of the file might simply been deleted and your script already supports filesystem options.

Carsten

[phil cook]

Your right Carsten. The dummy script created does contain those examples re mount options. Confused as to why then the script that mounts as per defined in this file does not contain any reference to mount options:

lv_mount()
{
j=$1
while [ $j -lt $(($1 + $2)) ]
do
mount ${LV[$j]} ${FS[$j]} >/dev/null 2>&1
if [ $? -ne 0 ]
then
message "ERROR: Failed to mount ${FS[$j]}"
j=$(($j - 1))
while [ $j -ge 0 ]
do
umount ${FS[$j]} >/dev/null 2>&1
if [ $? -ne 0 ]
then
message "ERROR: Failed to unmount file system ${FS[$j]}"
fi
j=$(($j - 1))
done
return 1
fi
j=$(($j + 1))
done
return 0
}

Has this installation been messed with in some way? It seems from this that should I change the pkg config script to include options, the mount script will ignore them anyway?

Phil C

[melvyn burnard]

My guess is thaht the scripts were done on an older version of SG, and then the SG version was upgraded without merging in the newer script options (as is normal ;-{ )
So you coul dprobably just recreate your scripts now and merge in the current information while tajking the opprotunity to use the new functionality

All good fun, they say

[Carsten Krege]

The good news is that you don't need to update then. The bad news is that your current package script is from an ooooooold SG version. Probably the cluster has been updated earlier and the scripts of the old version were simply re-used (which is ok).

The best is to move your package control file to the new script versions, you get with "cmmakepkg -s".

Carsten

[phil cook]

Ok chaps thanks very much for your advice, it's been very useful. I will try & regenerate the scripts on my uat node.

Best Rgds

Phil C

[phil cook]

I've now tried regenerating these scripts & although it's a little laborious copying in all your package settings it's worked first time.

Are there any other elements besides package control & config scripts that I should update for consistency?

Many thanks

[Carsten Krege]

No, not for consistency.

The unfortunate thing is that SG does not separate the base package control script from the customizable parts of the script. This way the base script could be easily patched and customers would not need to incorporate updates/changes/fixes of the base script by cut&paste.

Carsten