Dirty Region Logs for LSM mirrored volumes

[Thomas J. Harrold]

I have two systems, and two EMA12k arrays in two separate computer rooms.

Focusing on the storage: I'd like to mirror my data, using LSM, between the two storage arrays.

I believe that the "defaults" are not optimal. I think I want a DRL on each EMA array. (default seems to be create only one DRL, which means that it will reside on only one EMA array - failure of that array will nuke my DRL)

Here's what I think I need to do:

1) on each EMA array, present two LUNs to each host. I'll simplify:
hosta: dsk1 (50GB) dsk2 (1GB) dsk11 (50GB) dsk12 (1GB)
hostb: dsk1 (50GB) dsk2 (1GB) dsk11 (50GB) dsk12 (1GB

2) Initialize disks under LSM:
voldisksetup -i dskn (repeat for each disk)

3) Build disk group
voldg init dsk1 dsk2 dsk11 dsk12 nconfig=4

4) Build volume
volassist -g make 50g dsk1

5) Add second (mirror) plex:
volassist -g mirror dsk11

6) Add DRL plex:
volassist -g addlog dsk2

7) Add 2nd DRL plex:
volassist -g addlog
Am I missing something?

Thanks in advance.

[Han Pilmeyer]

There's best practice about LSM on multiple arrays in the documentation section of the Tru64 UNIX web site. I suggest you read that and then see if there are any additional questions.

With V5, LSM automatically allocates the configuration copies. They may all end up on the same array, which leaves you with an unusable configuration should that array fail. The best practice describes how to deal with this.

BTW. Note that the use of LSM for Disaster Tolerance is NOT supported by HP. You didn't say that that is the goal that you're trying to achieve, but I just want to be clear about this.

[Thomas J. Harrold]

I believe I have read all of the relevent docs, but If you care to point me to something I may have missed....

And I disagree with your statement about LSM and disaster tolerance. Although LSM has many limitations with TruCluster, it is supported for what we are trying to do.

[Han Pilmeyer]

I just hope you read this: http://h30097.www3.hp.com/docs/best_practices/BP_LSMHSGCONFIG/TITLE.HTM

I did try to say that what you are trying to do is supported. However using it for Disaster Tolerance is definitely NOT supported. The difference is subtle, but very important. Disaster Tolerance implies protecting against cascading failures. Typically people do this by introducing distance into the configuration. However introducing distance, by itself, does not offer Disaster Tolerance.

In other words, the configuration you use is supported. But if your intention is to protect yourself against the loss of data in the event of a disaster, then we suggest that you consider a form of hardware replication with Tru64 UNIX (such as EVA/CA, XP/CA or HSG80 DRM).

[Thomas J. Harrold]

Thanks for the clarification. Now I understand what you are getting at.

Let me take a look at the document you referenced.

-tjh