EMC, DR, SRDF, LVM

[Pedro Cirne]

Hi,

Here in our site we have a SAN with one Symmetrix that is replicated to another using SRDF.

I was wondering if someone knows if HP(SecurePath?) or EMC(PowerPath?) have any product to allow SRDF paths beeing defined as alternate links(or something similar) to PV, or any other solution that wil automatically recognize SRDF paths when a failover happens.

May be a "virtual" device with both paths?

Any ideias?

Thks!

Pedro

[Pedro Sousa · Accepted solution]

Yo Pedro.
ta-se?
nao sabia que andavas nisto ;). eu ja nao ando por aqui ha mttttt tempo.
mas ja agora que te encontrei por acaso aqui vai a minha resposta:

I think you don't want to run SRDF as alternate links as it would go against data integrity if it worked.

The target volumes are/should be write disable to keep you from writing on both EMCs at the same time.
here is a thing I found on the web some time ago that might explain it better:

SRDF Logical volume states

Volumes can be in three possible states.

* Not ready (NR) â canâ t be accessed by the host at all
* Write Disabled (RO) â can be accessed by the host for read only
* Write enabled (RW) â can be accessed by the host for read and write

The actual status of a volume depends on its SRDF state, and its Channel interface state. A Source volume has six different possible combinations of states, and a target volume has nine.

The desirable state for a source volume is SRDF state=RW and CI state=RW so volume state=RW If a primary volume CI state is RW, but the SRDF state is NR, then it may be possible to access the data from the target volume, if it is in the correct state.

The desirable state for a target volume is SRDF state=RO and CI state=RW so volume state=RO

fica bem ;)

[Stuart Abramson]

Pedro:

I don't exactly understand your question. What are you trying to achieve? What kind of "failover" are you talking about? Are the R1s and R2 both presented to the same server or to different servers? Are they in an MC/ServiceGuard Cluster?

In the normal course of events, that I'm familiar with, the R1s are presented to Server A, in Pittsburgh, for example, and the SRDF R2s are presented to Server B in Tennessee, 500 miles away. There would not be any PowerPath connections.

[Pedro Cirne]

Olá Pedro,

Tudo bem pela Suiça? Pois ás vezes venho aqui aos foruns para ter umas ideias :) Já te dei 10 pontos!

Agora em inglês:

What I need is to have things configured in a way that if the production storage fails, I can easely bring up the DR volumes and have servers using the DR storage.

Pedro, I know the volumes can have the three states you've mentioned, but in the event of a major failure on the production storage, I will have to manually reconfigure on all servers, all the LVM information with the new device files, because volumes on DR storage will have different /dev/dsk/c#t#d#, don't you agree?

Are there any alternatives?

Thks!

Pedro

Pedro, aguardo o teu contributo!

[Florian Heigl (new acc)]

the SRDF links in itself are only used for the real-time syncronisation between the symmetrixes.
PowerPath will give You real loadbalancing and better availability when using multiple paths to the LUNs presented by the symmetrix.

usually, You'd see four paths per LUN:
hba1 to symmetrix1
hba1 to symmetrix2
hba2 to symmetrix1
hba2 to symmetrix2
(symmetrix 1 is called R1 if I remember that correctly, and until it fails, all accesses will go to this one; in case of failure accesses will be re-routed to the other symmetrix, R2)

PowerPath will double the available bandwidth to the active symmetrix.
SecurePath does the very same thing.

From my point of view the difference between the two is such that HP won't support PowerPath and EMC^2 won't support SecurePath. :)

florian

[Pedro Sousa]

:)
I agree with what Stuart wrote.
what's the environment you have? :) I haven't been there for some time now.
do you have server A with R1s and server B with R2s?

whenever you want to use the R2s you'll have to change their state to writable and to configure them in Server B.

se quiseres falamos amanha melhor por telefone.

vou sair agora.
abraco,
Pedro
PS: nesta nao mereco tantos pontos :)

[Tim Nelson]

As mentioned the SRDF from one array to another is not instant sync (QOS) involved.

The second symm can be attached and have connectivity to the server in the event of a failure. The next trick is how to get the second symm into the UX config for failure but only in the event of a failure.

There may be policy settings that can be made in powerpath. I.e. never use this path unless specifically forced.

The PV link is another option as that device will not be used except for failover.

The issue still stands with data integrity on the second SRDF'd symm. Talk with your EMC rep or browse the EMC knowledge base or whitepapers.

I would have thought this would have been discussed way before purchasing the second symm or SRDF SW.

[Rita C Workman]

This is my configuration, so here goes.
Pedro,

First...SRDF are the data communication links (always redundant) that run between your sites. In fact they could both go down (I know I've had it happen)...this is what keeps your arrays communicating and sync'ing the data. Both sites are still up, just not talking to each other. Once the links are restored data begins humming across till it all catches up.

Next...PowerPath is the software (which I highly recommend) to allow for full utilization of your disks. It lets you use both paths/links off your hba to basically double your I/O to disk. Wonderful tool-easy to install and configure. It's strictly a local array thing-nothing remote involved here.

Now let's talk sync and DG and R1/R2. First alot depends on how your DG is set up. If it's sync, then it's mirrored; if it's adaptive, then your only as good as your Oracle logs and the last time your refreshed; if it's assync, then your going to be a little out of sync and if you were smart you set your Oracle logs to full sync to get you back ok.
Now, you set up your vg's like normal with the kind of disks you need. For remote sync replication your going to use R1/R2 disks (R1 primary and R2 your remote disks). Your going to figure which R1 disks you want and create your VG(s)with them. Next your going to create your DG or device group and add these R1 disks to the device group. The R2 disks will automatically be associated to that device group, because they are connected via the EMC/symm technology on each disk. Try running " symrdf list " and you'll see in column one the R1, then next column shows what R2 is attached to it...see my point. EMC does the mirroring behind the scenes.
Now I'm going to step out and not talk about setting up this at the R2-Remote site and host.
Like you said...only your R1 primary disk array goes down, so how does it get the R2 information...
...... Remember, they are just another mirror copy. So if one mirror goes down, the other is still up, sync'ing via SRDF links until the moment it went down.
But since they are R2 disks, you will need to run a command to make them (the R2 disks) RW..check out the command :
symrdf -g failover
This command will change the direction and make the remote site disks enabled. [...And now your disks are basically going to require some special attention when the R1 array comes back up. There are two commands you could use..failback or update. Failback of the device group will cause every track to be read (takes forever), but update just updates the R1 with changes from the R2 (may be preferred)... so it would look like:
symrdf -g update...]

Your app may feel the pinch coming back across the network running off those R2's, but if you look at your DG (device group) right after you create it, you will note it shows the disks as RW at both sites.

..and one last thing...the potential of an array going down is slim, but that is why I personally don't put my boot disk(s), or anything vg00 on them.

Whew....I'm tired now and late for a mtg...hope there's no typos or oopss...
Rgrds,
Rita

[Pedro Cirne]

Hi Rita,

Thank you for your reply.

I already understood what is SRDF and what are R1 and R2 volumes.

All that configuration is made by ECM engineers.

You talk about DG configuration, what is this? Is this EMC stuff?

What I don't understand is how I can "inform" LVM that the primary storage has failed and that the DR storage must be used.

Thks!

Pedro

[Rita C Workman]

If your replicating with SRDF...don't you have Solution Enabler loaded on your UNIX hosts ?
This is the EMC software that controls what your disks do/are. You have it if you can run any of the symcli commands like:
symrdf, symmir, symdev, symld...etc..

The older version of Solution Enabler was called WideSky. Just try running:

symcli

..see if it comes back with a version name & number..

Rita

[Pedro Cirne]

Hi Rita,

Yes I have Solution Enabler installed.

Pedro

[Hugo Cruz López]

Hi, Pedro !!

When we talk about powerPath, we are talking at a host level and if we talk about SRDF this es managed by the Symmetrix box, it has nothing to do about PowerPath.

Regards !!!