topic Re: Checking extends distribution of Logical Volumes against Physical Volumes in Operating System - Linux

Checking extends distribution of Logical Volumes against Physical Volumes

Eric SAUBIGNAC — Thu, 23 Feb 2017 17:05:08 GMT

Bonjour,

I desperately search how to see how the extends of a Logical Volume are distributed on the disks that make up the Volume Group. Especially when the Logical Volume is mirrored between 2 different storage array.

The idea behind this check is :

- to detect if a logical volume is mirrored on two volumes from a same array
- to verify wich extends are not synchronized

I had a case where one of my client configured a volume group with several disk from a given array, several from a second array, and has created a logical volume mirrored on the same array ... :-(

Under HP-UX we can simply control the localisation of extends of a logical volume whith lvdisplay and -v option. It doesn't work under Linux.

Exemple with vg01 which owns 2 disks from an array (disk12 en disk23) and 2 disks from another array (disk114 and disk142)

root@kskckca:/# lvdisplay -v /dev/vg01/lvol1
--- Logical volumes ---
LV Name                     /dev/vg01/lvol1
VG Name                     /dev/vg01
LV Permission               read/write
LV Status                   available/syncd
Mirror copies               1
Consistency Recovery        MWC
Schedule                    parallel
LV Size (Mbytes)            69988
Current LE                  17497
Allocated PE                34994
Stripes                     0
Stripe Size (Kbytes)        0
Bad block                   NONE
Allocation                  strict
IO Timeout (Seconds)        default

   --- Distribution of logical volume ---
   PV Name                 LE on PV PE on PV
   /dev/disk/disk12        17497    17497
   /dev/disk/disk114       17494    17494
   /dev/disk/disk23            3        3

      --- Logical extents ---
   LE       PV1                PE1      Status 1 PV2                PE2      Status 2
   00000000 /dev/disk/disk12   00000000 current /dev/disk/disk114 00003821 current
   00000001 /dev/disk/disk12   00000001 current /dev/disk/disk114 00003822 current
   00000002 /dev/disk/disk12   00000002 current /dev/disk/disk114 00003823 current
   .../...
   00017494 /dev/disk/disk12   00017494 current /dev/disk/disk23   00006241 current
   00017495 /dev/disk/disk12   00017495 current /dev/disk/disk23   00006242 current
   00017496 /dev/disk/disk12   00017496 current /dev/disk/disk23   00006243 current

Here I can see that almost all extends of this LV are mirrored between disk12 (my first test array) and disk114 (a volume from a second array). OK.
BUT the last 3 extends of the LV are mirrored between disk12 and disk23 ... that reside on the same storage array. Not good :-(

I can also see that all extends are synchronised : status = current

Is there anyway to get those useful informations under Linux ? I currently work with a Red Hat 7.3

Many thanks in advance

Eric

Re: Checking extends distribution of Logical Volumes against Physical Volumes

simplylinuxfaq — Tue, 28 Feb 2017 09:12:08 GMT

There is "-m" switch available with "lvdisplay" and "pvdisplay" which would show up the mapping of logical extents to physical volumes and physical extents. This may help you.

Example:

[root@ansible-host ~]# lvdisplay -m /dev/datavg/datalv
--- Logical volume ---
LV Path /dev/datavg/datalv
LV Name datalv
VG Name datavg
LV UUID eMOIpm-0iHa-gWHH-itPq-6lYA-fmVg-uhmYLH
LV Write Access read/write
LV Creation host, time ansible-host.example.com, 2016-12-28 04:53:13 -0500
LV Status available
# open 1
LV Size 296.00 MiB
Current LE 74
Segments 1
Allocation inherit
Read ahead sectors auto
- currently set to 256
Block device 253:3

--- Segments ---
Logical extents 0 to 73:
Type linear
Physical volume /dev/sdb1
Physical extents 0 to 73

[root@ansible-host ~]# pvdisplay -m /dev/sdb1
--- Physical volume ---
PV Name /dev/sdb1
VG Name datavg
PV Size 300.00 MiB / not usable 4.00 MiB
Allocatable yes (but full)
PE Size 4.00 MiB
Total PE 74
Free PE 0
Allocated PE 74
PV UUID cjcgSN-tzcq-DMKn-fJ0R-NE9o-qhig-VZDjKY

--- Physical Segments ---
Physical extent 0 to 73:
Logical volume /dev/datavg/datalv
Logical extents 0 to 73

Re: Checking extends distribution of Logical Volumes against Physical Volumes

Eric SAUBIGNAC — Tue, 28 Feb 2017 11:25:07 GMT

Thank you for the information. I totally missed the -m / '--maps' option. I gave a kudo for this.

Unfortunatly, I am afraid the informations given by the "-m" are very light regarding a mirrored raid1 logical_volume: it gives some details about the "internal" Logical Volumes but I still don't know how a given logcal extend is mapped to physical extent(s).

To clarify, here is a small test I did:

- created a vg with 2 x 10 GB san virtual volumes from a given storage array, and 2 other 10 GB from an other array

- tagged 2 virtual volumes from one array with BDX, and the 2 others with LAC :

# pvs -o name,vg_name,tags
PV VG PV Tags
/dev/mapper/YRO_BDX_POCSGLXPK10_01 vg_sglxpk10 BDX
/dev/mapper/YRO_BDX_POCSGLXPK10_02 vg_sglxpk10 BDX
/dev/mapper/YRO_LAC_POCSGLXPK10_01 vg_sglxpk10 LAC
/dev/mapper/YRO_LAC_POCSGLXPK10_02 vg_sglxpk10 LAC

- then, I tried to create a 11GB mirrored volume. Because the size is greater than one san volume, the logical volume will lie on 2 san volumes and the mirror on the other 2. The challenge was to have all extends from one source (for example BDX) mirrored on the other source (LAC). So I tried this :

lvcreate -n lvol1 -L 11G -m 1 --type raid1 vg_sglxpk10 @bdx @LAC

If I try to analyze how extends are distributed against physical volumes with the –m option:

# lvdisplay -m /dev/vg_sglxpk10/lvol1
.../...
--- Segments ---
Logical extents 0 to 2815:
    Type                raid1
    Monitoring          monitored
    Raid Data LV 0
      Logical volume    lvol1_rimage_0
      Logical extents   0 to 2815
    Raid Data LV 1
      Logical volume    lvol1_rimage_1
      Logical extents   0 to 2815
    Raid Metadata LV 0 lvol1_rmeta_0
    Raid Metadata LV 1 lvol1_rmeta_1

I can see the distribution against "internal" Logical Volume lvol1_rimage_0 and lvol1_rimage_1, not directly against Physical Volumes. So I am unable to qualify if the mirroring is done between 2 volumes from the same array [ bad :-( ]or from a different array [ nice :-) ]

And in the case of this test there is a real problem:

- First, _rimage_0 lies on 1 volume from an array and 1 volume from the other array. Same thing for the internal lvol _rimage_1. It means that a logical extend in this configuration could be mirrored between a physical extent from one storage and an other one from the same array.

It can be checked like this:
# lvs -a -o name,vg_name,devices vg_sglxpk10
LV VG Devices
lvol1 vg_sglxpk10 lvol1_rimage_0(0),lvol1_rimage_1(0)
[lvol1_rimage_0] vg_sglxpk10 /dev/mapper/YRO_BDX_POCSGLXPK10_01(1)
[lvol1_rimage_0] vg_sglxpk10 /dev/mapper/YRO_LAC_POCSGLXPK10_02(0)
[lvol1_rimage_1] vg_sglxpk10 /dev/mapper/YRO_BDX_POCSGLXPK10_02(1)
[lvol1_rimage_1] vg_sglxpk10 /dev/mapper/YRO_LAC_POCSGLXPK10_01(0)
[lvol1_rmeta_0] vg_sglxpk10 /dev/mapper/YRO_BDX_POCSGLXPK10_01(0)
[lvol1_rmeta_1] vg_sglxpk10 /dev/mapper/YRO_BDX_POCSGLXPK10_02(0)

- Secondly, I can see from a vgdisplay the global allocation of physical extents. And it clearly shows that both volumes from the "BDX" location are full. So probably the mirroring is done between those 2 volumes for the first 2556 extends, and the two volumes from the array located @ "LAC" for the last 260 extends. Extract :

# vgdisplay -v vg_sglxpk10
--- Physical volumes ---
PV Name /dev/mapper/YRO_BDX_POCSGLXPK10_01
Total PE / Free PE 2556 / 0

PV Name /dev/mapper/YRO_BDX_POCSGLXPK10_02
Total PE / Free PE 2556 / 0

PV Name /dev/mapper/YRO_LAC_POCSGLXPK10_02
Total PE / Free PE 2556 / 2295

PV Name /dev/mapper/YRO_LAC_POCSGLXPK10_01
Total PE / Free PE 2556 / 2295

In a summary I can say there is a problem but I can't analyze it in further details … that's my problem. And the case presented here is rather simple but could be more complex in "real world" IT.

Eric

Re: Checking extends distribution of Logical Volumes against Physical Volumes

simplylinuxfaq — Tue, 28 Feb 2017 13:02:31 GMT

Since it is known where the data has to be dumped, can't we specify the luns/pvs directly while creating an lvm.

Hint

[root@ansible-host ~]# lvcreate -v --extents 100 -n testlv testvg -m 1 --type raid1 /dev/sdf1 /dev/sdh1
Using volume group(s) on command line.
Archiving volume group "testvg" metadata (seqno 13).
Creating logical volume testlv
Creating logical volume testlv_rimage_0
Creating logical volume testlv_rmeta_0
Creating logical volume testlv_rimage_1
Creating logical volume testlv_rmeta_1
activation/volume_list configuration setting not defined: Checking only host tags for testvg/testlv_rmeta_0.
Creating testvg-testlv_rmeta_0
Loading testvg-testlv_rmeta_0 table (253:7)
Resuming testvg-testlv_rmeta_0 (253:7)
Clearing metadata area of testvg/testlv_rmeta_0
Initializing 512 B of logical volume "testvg/testlv_rmeta_0" with value 0.
Removing testvg-testlv_rmeta_0 (253:7)
activation/volume_list configuration setting not defined: Checking only host tags for testvg/testlv_rmeta_1.
Creating testvg-testlv_rmeta_1
Loading testvg-testlv_rmeta_1 table (253:7)
Resuming testvg-testlv_rmeta_1 (253:7)
Clearing metadata area of testvg/testlv_rmeta_1
Initializing 512 B of logical volume "testvg/testlv_rmeta_1" with value 0.
Removing testvg-testlv_rmeta_1 (253:7)
Creating volume group backup "/etc/lvm/backup/testvg" (seqno 15).
Activating logical volume "testlv" exclusively.
activation/volume_list configuration setting not defined: Checking only host tags for testvg/testlv.
Creating testvg-testlv_rmeta_0
Loading testvg-testlv_rmeta_0 table (253:7)
Resuming testvg-testlv_rmeta_0 (253:7)
Creating testvg-testlv_rimage_0
Loading testvg-testlv_rimage_0 table (253:8)
Resuming testvg-testlv_rimage_0 (253:8)
Creating testvg-testlv_rmeta_1
Loading testvg-testlv_rmeta_1 table (253:9)
Resuming testvg-testlv_rmeta_1 (253:9)
Creating testvg-testlv_rimage_1
Loading testvg-testlv_rimage_1 table (253:10)
Resuming testvg-testlv_rimage_1 (253:10)
Creating testvg-testlv
Loading testvg-testlv table (253:11)
Resuming testvg-testlv (253:11)
Monitoring testvg/testlv
Wiping known signatures on logical volume "testvg/testlv"
Initializing 4.00 KiB of logical volume "testvg/testlv" with value 0.
Logical volume "testlv" created.
[root@ansible-host ~]# pvdisplay -m /dev/sd{f,g,h,i}1
--- Physical volume ---
PV Name /dev/sdf1
VG Name testvg
PV Size 1023.00 MiB / not usable 3.00 MiB
Allocatable yes
PE Size 4.00 MiB
Total PE 255
Free PE 154
Allocated PE 101
PV UUID 4vjqDE-I8d8-Uo86-Kxiq-YoFh-Djys-gyfQuF

--- Physical Segments ---
Physical extent 0 to 0:
Logical volume /dev/testvg/testlv_rmeta_0
Logical extents 0 to 0
Physical extent 1 to 100:
Logical volume /dev/testvg/testlv_rimage_0
Logical extents 0 to 99
Physical extent 101 to 254:
FREE

--- Physical volume ---
PV Name /dev/sdg1
VG Name testvg
PV Size 1023.00 MiB / not usable 3.00 MiB
Allocatable yes
PE Size 4.00 MiB
Total PE 255
Free PE 255
Allocated PE 0
PV UUID sVkMc4-E8KT-GfYr-q0Nb-9BAd-cvUD-a7u7TW

--- Physical Segments ---
Physical extent 0 to 254:
FREE

--- Physical volume ---
PV Name /dev/sdh1
VG Name testvg
PV Size 1023.00 MiB / not usable 3.00 MiB
Allocatable yes
PE Size 4.00 MiB
Total PE 255
Free PE 154
Allocated PE 101
PV UUID 4K9iwh-hrfW-75Ce-x6nD-hEa2-FA1F-RcwDHQ

--- Physical Segments ---
Physical extent 0 to 0:
Logical volume /dev/testvg/testlv_rmeta_1
Logical extents 0 to 0
Physical extent 1 to 100:
Logical volume /dev/testvg/testlv_rimage_1
Logical extents 0 to 99
Physical extent 101 to 254:
FREE

--- Physical volume ---
PV Name /dev/sdi1
VG Name testvg
PV Size 1023.00 MiB / not usable 3.00 MiB
Allocatable yes
PE Size 4.00 MiB
Total PE 255
Free PE 255
Allocated PE 0
PV UUID AGTOs1-Mlba-favG-2xV9-bOzl-jVOm-3lmJXY

--- Physical Segments ---
Physical extent 0 to 254:
FREE

Re: Checking extends distribution of Logical Volumes against Physical Volumes

Eric SAUBIGNAC — Sun, 19 Mar 2017 12:01:01 GMT

Hi,

Sorry for this late answer.

I agree with you : if the job is correctly done @ creation time, well ... the job is done.

But, my problem is mostly later in the time : "what, if the initial job has not been done thoughtfully ?" Not sure that existing tools are enough to analyzes & corrects problems

Eric