analyze-my-lvm doesn't work for more than "1" /dev/mapper/* lvm PV devices on RHEL6.2

[bzium]

Function GiveMapperOfdm() in analyze-my-lvm doesn't work for more than "1" /dev/mapper/* lvm PV devices on RHEL6.2

/dev/mapper/(dev_name) is symbolic link to /dev/dm-X "stat" command in GiveMapperOfdm should be run with "-L" switch.

GiveMapperOfdm () {

major=`stat -c "%t" $1 2> /dev/null`
minor=`stat -c "%T" $1 2> /dev/null`

for i in `ls /dev/mapper/*`; do
        mj=`stat -c "%t" $i`
        mn=`stat -c "%T" $i`
        if [ "$mj" = "$major" ] && [ "$mn" = "$minor" ]; then
                echo "$i"
                return
        fi
done
echo $1
}

should be

GiveMapperOfdm () {

major=`stat -L -c "%t" $1 2> /dev/null`
minor=`stat -L -c "%T" $1 2> /dev/null`

for i in `ls /dev/mapper/*`; do
        mj=`stat -L -c "%t" $i`
        mn=`stat -L -c "%T" $i`
        if [ "$mj" = "$major" ] && [ "$mn" = "$minor" ]; then
                echo "$i"
                return
        fi
done
echo $1
}

[Bruno Cornec]

I'm not sure I understand your problem. stat doesn't use -L as we need to have all the links, and the role of GiveMapperOfdm is to just give the /dev/mapper/ device correspondance to the /dev/dm-* one.

So while I'm not disagreeing that there may be an issue with multiple PVs, I'd like to get the /var/log/mondoarchive.log to have a better understanding of the issue, and see how to best fix it. For the moment, I think your fix risks to create more issues elsewhere.

TIA, Bruno.

[bzium]

Please take givemappefdm.sh.gz from attachment and check two or more devices.

# ls -la /dev/mapper/mpath*
lrwxrwxrwx 1 root root 7 Apr 18 10:34 /dev/mapper/mpatha -> ../dm-2
lrwxrwxrwx 1 root root 7 Apr 18 10:34 /dev/mapper/mpathb -> ../dm-3
lrwxrwxrwx 1 root root 8 Apr 19 10:27 /dev/mapper/mpathc -> ../dm-10
# ./givemappefdm.sh /dev/mapper/mpatha
/dev/mapper/mpatha
# ./givemappefdm.sh /dev/mapper/mpathb
/dev/mapper/mpatha
# ./givemappefdm.sh /dev/mapper/mpathc
/dev/mapper/mpatha

Your function always display first device /dev/mapper/mpatha. After that try givemappefdm-new.sh.gz from attachments.

# ./givemappefdm-new.sh /dev/mapper/mpatha
/dev/mapper/mpatha
# ./givemappefdm-new.sh /dev/mapper/mpathb
/dev/mapper/mpathb
# ./givemappefdm-new.sh /dev/mapper/mpathc
/dev/mapper/mpathc

If you do not add "-L" stat for each device will show:

# stat -c "%T" /dev/mapper/mpath*
0
0
0
[root@proddb1 mondo-test]# stat -c "%t" /dev/mapper/mpath*
0
0
0

All devices have a "major minor = 0 0". If you add "-L" switch stat show correct "major minor".

# stat -L -c "%T" /dev/mapper/mpath*
2
3
a
# stat -L -c "%t" /dev/mapper/mpath*
fd
fd
fd

Do you understand now what's the problem?

analyze-my-lvm always take first /dev/mapper/ device in this case show /dev/mapper/mpatha. /dev/mapper/mpatha is not a physical device for LVM

# pvs
  PV                 VG         Fmt  Attr PSize    PFree
  /dev/mapper/mpathb vg_puedata lvm2 a--     2.00g 104.00m
  /dev/mapper/mpathc vg_test    lvm2 a--  1020.00m      0
  /dev/sda2          vg01       lvm2 a--    17.80g   2.30g

# ./analyze-my-lvm
/sbin/lvmdiskscan
/sbin/lvm
i=2
LVM version >= 2.0 found.
Just before you extrapolate mountlist to include RAID partitions,
extrapolate it to include the following LVM drives and partitions:-
>>>>> /dev/mapper/mpatha /dev/mapper/mpatha /dev/sda2
To get started, type:-
(insmod dm-mod)
(insmod dm_mod)
# lvm vgchange -an
# echo y | lvm pvcreate -ff /dev/mapper/mpatha
# echo y | lvm pvcreate -ff /dev/mapper/mpatha
# echo y | lvm pvcreate -ff /dev/sda2
# lvm vgscan

And with "-L" switch.

# ./analyze-my-lvm
/sbin/lvmdiskscan
/sbin/lvm
i=2
LVM version >= 2.0 found.
Just before you extrapolate mountlist to include RAID partitions,
extrapolate it to include the following LVM drives and partitions:-
>>>>> /dev/mapper/mpathc /dev/mapper/mpathb /dev/sda2
To get started, type:-
(insmod dm-mod)
(insmod dm_mod)
# lvm vgchange -an
# echo y | lvm pvcreate -ff /dev/mapper/mpathc
# echo y | lvm pvcreate -ff /dev/mapper/mpathb
# echo y | lvm pvcreate -ff /dev/sda2
# lvm vgscan

[Bruno Cornec]

Many thanks for your explanations, that showed I completely missed the point :-(.

Now fixed in rev [3001]. Will be in mindi 2.1.2.