Migration HP EVA to 3PAR using MPX on SUSE Linux 10,11

[TWRockstar]

Hello friends,

Need your valuable ideas and suggestions.

We are planning to migrate from EVA to 3PAR using mpx router offline migration.I am not a storage expert so i need some help with this. After long research on several websites including SPOCK i got to know that currently our linux servers are running native multipathing and it is supported by 3PAR so on linux side we don't need to un-install before migration and reinstall MPIO after migrating to 3PAR as on Windows 2003 we uninstall and re-install.

Currently my configuration:
Present sorage array is HP EVA (source) and planning to migrate to 3PAR(destination array)
SUSE Linux version 10 and patchlevel 4 x86_64
Using Native multipathing.
HBA emulex with port speed 4gbit
device-mapper-1.02.13-6.18.7
multipath-tools-0.4.7-34.66.1
qlogic-firmware-1.0-13.23.1

We performed host based migration but this array based migration like block to block so we are new to this kind of migration.

I thought like before migration start i will update my multipath configuration file with 3PAR device as recommended and then umount filesystem(commenting the storage path in /etc/fstab), shutdown the host. After migration  completes poweron the host start the multipath and uncomment the storage paths in /etc/fstab, mount the filesystem.

Are these correct steps and do we need to perform any additional task. Any help is highly appreciated.

Thanks

Tim Wright.

[Dennis Handly]

>using mpx router offline migration

You don't want to use online Peer motion?

[TWRockstar]

Dennis,

We are not using online peer motion.

We are just using MPX router where on EVA unpresent the host and then  present it to MPX and Destination 3PAR host are presented to MPX then we perform block level data migration. Once the migration is completed we unpresent the Destination 3PAR host from MPX and present it to host again. Zone them and then bring the servers up.

Can you please let me know whether my procedure is correct with respect updating multipath config file, unmounting filesystem, bring the server down. After migration bring the server up, start multipath, mount the filesystem and anything else i need to follow apart from this.

Thanks,

Tim

[TWRockstar]

Any thoughts or suggestion please. Looking forward for your valuable words on this issue.

Thanks,
Tim

[Matti_Kurkela]

If you are using LVM, you can present the new disks in addition to the old ones, then use either pvmove (well tested) or LVM mirroring (not so well known) to migrate the data without downtime.

With pvmove, the general procedure for each VG to be migrated is:

- present enough disks/LUNs from the new storage to cover the VG (the number and size of the disks will not need to match the old ones: the only requirement is that the total size is enough)

- add new disks as new PVs to the VG (pvcreate, then vgextend)

- use pvmove to move the data from each old PV to one or more of the new ones

- use vgreduce to remove the now-empty old PVs from the VG

- use multipath -f <multipath-device> to unmap the old multipath devices (or use "multipath -F" to clear all unused multipath devices at once)

- clean up the old disk devices at the OS level, (for example, "echo 1 > /sys/block/sdX/device/delete" to tell the system that /dev/sdX will soon be intentionally removed and the OS should stop accessing it)

- unpresent old disks at the SAN level

If you are not using LVM, then there is no general solution for online migration: your online migration possibilities will depend on what is on the disks. (Oracle ASM: maybe yes, application filesystems: usually no, unless the application specifically supports it)

[TWRockstar]

Thanks MK. From your solution i see there is no use of MPX as we will be presenting new disks directly to the host from the 3par and performing usual approach of pvmove or lvm mirrioring. Is there any way we can use MPX during the online migration instead of performing pvmove or lvmmirroring ?