LVM mirroring across sites

[Srinikalyan]

We need to migrate around 20TB of XP storage array data from one site to EVA storage of another site which is seperated by around 100KM. we have a dedicated 2Gbps DWDM link for the transfer. Can I do the LVM mirroring from one site HP server to another site HP server for this purpose.
Thanks,
Srini

[Ganesan R]

Hi Srinikalyan,

Yes it is possible. As long as the remote luns are visible and accessible by the local host, you can mirror it.

[Emil Velez]

Might Continuous access be a possible solution.

[Avinash20]

Yes, it possible but it depends on the connectivity between the storages
With recent advances in Fibre Channel technologies, i.e., DWDM, the distances we are now talking about are up to 100Km


Also refer
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1315353

[Sani]

What data it is?
If it is database then you may use some database replication techniques.

Regards
Sani

[Srinikalyan]

Emil,
We do have only CA/XP licenses. I need to replicated the data from XP to EVA. So I am afraid that I can't use CA/XP technologies.

Sani,
This is for migrating Oracle data of around 30TB from production to performance environment. Since the database is huge I can't use oracle export/import and also dataguard because this is for initial db copy.
Do you have any other mechanism?

Regards,
Srini

[Sani]

Hi Srini ,

For initial setup i think the best option is backup restore. Further you can use oracle data guard to replicate the database.

I dont think CA will work between eva/xp.As per my knowledge some thrid party software is available for data replication.

Regards
Sani.

[Shahul · Accepted solution]

Hi Srini,

Hostbase Mirroring is possible as long as you can see the disks from both local and remote arrays. A couple of potential problems with that is,

1. Host base mirroring can impact the performance production, especially when the volume is 20 to 30 TB.

2. If your Volume Groups are created with a small number of MAx PV per VG, then you may not be able to double the amount of disk spac in Volume groups. This is a common problem.

The other options you have is, Backup and restore. You may already running backups every, you can take those to remote site and restore there, as it is for performance testing, I don't think you need to keep updating this.

if you already have a CA link between these two sites, then try to get the remote XP array visible to the remote host where EVA is present. Try host mirroring between remote XP and EVA, this will not impact production. But you may need to split CA link for sometime.

Regards
Shahul

[Ganesan R]

Hi Srinikalyan,

There is another option which depends on the possiblities..

If you worry about performance over WAN, you can get the remote storage locally, create VG's on new storage and copy the data onto it. Once copied export the VG's. You can import the VG's on remote server then.

Again, it is a option only.

[Steven E. Protter]

Shalom,

It is possible with the speed of that link.

If its full duplex, you can transfer the data in 10 or so hours, depending on the lag. Lag can be checked with ping time.

There will be a significant time hit as I/O occurs at both ends.

Any kind of line issues could disrupt the transfer.

It might be faster to shut down services, have file systems set up in both locations and do scp -rp site to site.

Transferring the data via tape backup, might be the most reliable way of doing it, even though it means driving the tapes 100 km.

I would leave several days for the transfer based on the difficulties of doing this with LVM

SEP

[TTr]

30TB is a lot of data. You need to evaluate the impact that server or storage mirroring might have on your production environment. The last thing you want to do is add more load to it. I would look into the backup restore system as the only option here. After you restore the data, you can even apply the recent logs (again from your backup system) to make the perf env current.

[Srinikalyan]

Thanks everyone for sharing your expertise knowledge. I will look into the possible solution within this.
Thanks,
Srini