[Linux-cluster] CLVM not activating LVs

[Christine Caulfield]

On 26/08/09 14:19, Jakov Sosic wrote:
> Hi! CLVM not activating logical volumes.
>
>
>
> I have a major issues with CLVM. It is not activating volumes in my
> VG's. I have 2 iSCSI volumes, and one SAS volume with 3 VG's. On
> node01, all logical volumes in 1 iSCSI and on SAS are activated. Other
> iSCSI - zero. On node02 same situation. On node03 only lv's from SAS
> are activated. lvm.conf is same on the machines....
>
> This is very strange, because when I boot the machines, all the
> services are stopped, so logical volumes shouldn't be activated.
>
> Here is the situation:
>
> [root at node01 lvm]# vgs
>    VG          #PV #LV #SN Attr   VSize  VFree
>    VolGroupC0    1   7   0 wz--nc  3.41T 1.48T
>    VolGroupC1    1   0   0 wz--nc  3.41T 3.41T
>    VolGroupSAS   1   2   0 wz--nc 20.63G 4.63G
>
> [root at node01 lvm]# lvs
>    LV            VG          Attr   LSize
>    nered1        VolGroupC0  -wi-a- 200.00G
>    nered2        VolGroupC0  -wi-a- 200.00G
>    nered3        VolGroupC0  -wi-a-   1.46T
>    nered4        VolGroupC0  -wi-a-  20.00G
>    nered5        VolGroupC0  -wi-a-  20.00G
>    nered6        VolGroupC0  -wi-a-  20.00G
>    nered7        VolGroupC0  -wi-a-  20.00G
>    sasnered0     VolGroupSAS -wi-a-   8.00G
>    sasnered1     VolGroupSAS -wi-a-   8.00G
>
> [root at node03 cache]# vgs
>    VG          #PV #LV #SN Attr   VSize  VFree
>    VolGroupC0    1   0   0 wz--nc  3.41T 3.41T
>    VolGroupC1    1   0   0 wz--nc  3.41T 3.41T
>    VolGroupSAS   1   2   0 wz--nc 20.63G 4.63G
>
> [root at node03 lvm]# lvs
>    LV          VG          Attr   LSize Origin
>    sasnered0   VolGroupSAS -wi-a-   8.00G
>    sasnered1   VolGroupSAS -wi-a-   8.00G
>
>
> here is my lvm.conf:
>
> [root at node01 lvm]# lvm dumpconfig
>    devices {
>    	dir="/dev"
>    	scan="/dev"
>    	preferred_names=[]
>    	filter=["a|^/dev/mapper/controller0$|",
> "a|^/dev/mapper/controller1$|", "a|^/dev/mapper/sas-xen$|", "r|.*|"]
>    	cache_dir="/etc/lvm/cache"
>    	cache_file_prefix=""
>    	write_cache_state=0
>    	sysfs_scan=1
>    	md_component_detection=1
>    	md_chunk_alignment=1
>    	ignore_suspended_devices=0
>    }
>    dmeventd {
>    	mirror_library="libdevmapper-event-lvm2mirror.so"
>    	snapshot_library="libdevmapper-event-lvm2snapshot.so"
>    }
>    activation {
>    	missing_stripe_filler="error"
>    	reserved_stack=256
>    	reserved_memory=8192
>    	process_priority=-18
>    	mirror_region_size=512
>    	readahead="auto"
>    	mirror_log_fault_policy="allocate"
>    	mirror_device_fault_policy="remove"
>    }
>    global {
>    	library_dir="/usr/lib64"
>    	umask=63
>    	test=0
>    	units="h"
>    	activation=1
>    	proc="/proc"
>    	locking_type=3
>    	fallback_to_clustered_locking=1
>    	fallback_to_local_locking=1
>    	locking_dir="/var/lock/lvm"
>    }
>    shell {
>    	history_size=100
>    }
>    backup {
>    	backup=1
>    	backup_dir="/etc/lvm/backup"
>    	archive=1
>    	archive_dir="/etc/lvm/archive"
>    	retain_min=10
>    	retain_days=30
>    }
>    log {
>    	verbose=0
>    	syslog=1
>    	overwrite=0
>    	level=0
>    	indent=1
>    	command_names=0
>    	prefix="  "
>    }
>
>
> Note that logical volumes form C1 were present on node01 and node02,
> but after the node03 joined cluster they dissapeared. I'm running
> CentOS 5.3.
>
> This is really dissapointing. Enterprise Linux? Linux maybe but not
> Enterprise... After much trouble with linux dm-multipath issues with my
> storage - which are unresolved and are waiting for RHEL 5.4, now clvmd.
>
> Note that locking (DLM), cman, rgmanager, qdisk and all the other
> cluster services are working without problems. I just don't get it why
> is CLVM behaving this way?
>
> I'm thinking about switching to non-clustered LVM - but are there
> issues with possible corruption of metadata? I won't create any new
> volumes nor snapshots or anything similar. Setup is done and it should
> work like this for the extended period of time.... But are there issues
> with activation or something else changing metadata?
>


You need to mark the shared VGs clustered using the command

# vgchange -cy <VGname>


If you created them while clvmd was active then this is the default. If 
not then you will have to add it yourself as above.


Chrissie