lvm and emc bcv

[Dee Conrad]

Hi All,

Really dumb question time...sorry. We have bcv's that are one/one that work correctly. One lun, one volume group, and one lv...equates to absolutely no problems.

Now we are trying to work with production volume groups and lv's on a different system. Again - the one/one works great, but when I try to setup bcv's for 2 logical volumes in the same volume group I get errors. What we is as follows:

1. umount, vgchange, and vgexport the bcv volumes
2. sync the bcv's with the source's using navisphere/naviseccli
3. fracture the source's/bcv's
4. Run vgchgid, mkdir, mknod, vgimport, vgchange, and vgcfgbackup

All is good to this point - however, when I go to run the fsck on the individual lvol's they both show as corrupt.

What has transpired with the source luns is that we have several filesystems/lvols that were orginally created as small lvols (I know this is small - 4gb), and then one of the lvol's was extended to 6gb. All of this was done before we created a bcv for each of these filesystems.

For testings - when I try to create the bcv for them - it shows corrupt in the fsck. If I create a new volume group, 2 new lv's, and then dd or cp the source information to the new test volumes, and link the test volumes to he bcv's then everything works correctly.

Anyway to correct this problem without having to recreate new lv's and volume groups? I already opened a ticket with EMC, and they said it was an lvm issue. Thanks.

ps - we are running emc clariion cx4-240 flare code 28, and hpux 11.11 systems.

[Duncan Edmonstone]

so when you split the BCVs are the filesystems mounted or not? you give the impression above that they're not...

also are all the LUNs in the volume group in a single EMC timefinder consistency group so that the aplit of all the LUNs in the VG is atomic?

HTH

Duncan

[Dee Conrad]

source luns/filesystems yes, but bcv is not mounted.

[Dee Conrad]

Sorry - didn't see the second question initially. I inherited some things, and so I am pretty new to this. On the emc side, I have 2 different clone groups, and no it isn't a consistency group.

Will check into that documentation, but any other help would be great.

Thanks.

[chris huys_4]

Hi,

Normally the only way to be sure that all the source data is "flushed" to "source" disk so that the bcv "target disk" "has all 'consistent' data", is through a stop of all applications on the "
"filesystems located on the source disk(s) ", unmount the 'source filesystem(s)
', then take the bcv, then remount back the 'source filesystem(s)'.

There is a "online workaround" of the above procedure and that is instead of the unmount do a remount with the filesystem(s) still mounted, which will also result in a complete 'flush' of all source filesystem data to "source" disk(s).

i.e.
# mount -o remount

NOTE: Not sure if the remount option is allready there in vxfs 3.5 (the latest vxfs version that is supported on HP-UX 11.11)

Greetz,
Chris