StoreVirtual Storage
1748239 Members
3680 Online
108759 Solutions
New Discussion юеВ

Lefthand/P4000 Managers

 
Panzer_1
Frequent Advisor

Lefthand/P4000 Managers

I am designing a P4000 solution. Storage nodes will be placed in site A and site B. Site A is the primary site, and site B will only be used if site A is taken out by an avalance. The correct way to do this is to use a Failover Manager in a third site that communicates with both site A and B, but a third site is not available in this case.

 

If I place a Failover Manager in site A, I will not be able to run my systems in site B after the avalance. If I place a Failover Manager in site B, I will risk loosing access to my storage in site A if the link between the sites goes down for a short period.

 

One possible solution could be to use a Virtual manager instead of a Failover Manager. The storage system will (as I understand it) go down, but can be started again in any location.

 

Any other suggestions?

 

 

P.S. This thread has been moved from Storage Area Networks (SAN) (Small and Medium Business) to HP StoreVirtual Storage / LeftHand. - Hp Forum Moderator

 

8 REPLIES 8
Steven Clementi
Honored Contributor

Re: Lefthand/P4000 Managers

It will depend upon the configuration of your cluster(s).  Are you planning on having a Multi-site Cluster? or 2 single clusters with Remote Copy schedules?

 

How many nodes per site?

Steven Clementi
HP Master ASE, Storage, Servers, and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5, vSphere 6.x)
RHCE
NPP3 (Nutanix Platform Professional)
Panzer_1
Frequent Advisor

Re: Lefthand/P4000 Managers

Only 1 node per site, configured with Network RAID10

oikjn
Honored Contributor

Re: Lefthand/P4000 Managers

just so you are aware, you are saying you want to do a multi-site cluster, but mention that you are going to run as an active and passive site... multi-site is WAY overkill for that situation because the second site is also active in that situation.  Better to use remote snapshots to the remote site as you don't have to worry about the FOM be as concerned about bandwidth and latency.  The site-site link is EXPENSIVE to actually run a multi-site cluster.  If you can afford that and NEED that, then you can afford to find a 3rd site either at a colo or in the cloud that can host the FOM.  If you can't, then you should re-examine the need for using a multi-site cluster and seriuosly consider just using remote snapshots...  there would be slight downtime associated with that, but it is minimal and if the cost of that downtime isn't in the upper 5 or 6 fixture range, its probably better to use that.

CalvinZito
HPE Blogger

Re: Lefthand/P4000 Managers

Totally agree with @oikjn.  If your run with FOM and two active sites, your running a multi-site SAN.  That's a great solution for near instantaneous failover but if you you want to have recovery (and longer RTO is ok), it will be less complicated to just do remote replication. 

 

However note that the FOM can run on just about any thing - it's a very lightweight FOM to break ties.  But does need to be in a third site or you risk split-brain.

Panzer_1
Frequent Advisor

Re: Lefthand/P4000 Managers

I am aware of this. However this is a design for a special need, let me try to explain:

The system is beeing used for monitoring the movements of a big rock on top of a mountain. Site A is the primary site for the guys doing the monitoring, and it is located at sea level. This site will probably be flooded when the rock hits the sea. Site B is located in a safe area on top of the mountain, connected to site A with a stable link. At some point all personell will be relocated to site B, and this i s why I need to be able to run site B even if site A and the link is permanently down.

oikjn
Honored Contributor

Re: Lefthand/P4000 Managers

So if I get this correctly, you are trying to replicate data as quick as possible off site for when the 2nd site actually will fail.

 

Definitely a special situation.

 

Presumably you are running fiber or wifi that will be able to handle the link speed requirements.  Keep in mind that the link will determin your maximum write speed to the storage so if the link is too slow your write performance will be too slow and you probably won't be able to store the data using the system.

 

Assuming the speed is sufficient and you are running the site-site link yourself, I wouldn't actually treat this like a multi-site cluster and just treat it like a simple cluster.  Ideal would be a FOM and VSA at the permanent site and a 2nd VSA at the temp site.  That way when the temp site goes offline the cluster would stay online and live.  You could use a virtual manager running on the permanent site VSA, but its better to run a FOM if you can.

 

The critical issue is going to be making sure you have your speed between sites to support the SAN load.  If you don't, you are pretty much screwed.

Panzer_1
Frequent Advisor

Re: Lefthand/P4000 Managers

Thank you for your answer. However, I dont't think have explained the sitation good enough.

 

The 2 P4000 systems should be in sync all the time, so the SAN load should not be an issue. The primary site A will at some point fail, and all personell will then move to site B. The problem is then to get the P4000 in site B to run when site A is down and we don't have a third location with a FOM. I could of course run the FOM in site B, but then the primary site S will go down if the link between the sites goes down temporarily.

 

Summary:

Site A should be running at all times until disaster stikes.

Site B must be running within hours after site A goes down permanently.

 

Is the Virtual manager a good solution here? As I understand it, it can be manualy started on either site when the administrator has decided which site is OK.

oikjn
Honored Contributor

Re: Lefthand/P4000 Managers

SAN load IS an issue.  you MUST have enough bandwidth between the two sites to handle 100% of the write capacity of the SAN as that link will likely be the limiting factor with your SAN latency and throughput capacity.

 

So...  assuming your link quality is sufficient for a multi-site san (you still haven't said it is yet), then it does sound like you could get away with running a SAN with a virtual manager running at Site A (since you want that one to always be up until the really goes dead) and then when it does go dead, you will have to force the virtual manager to start on the node at site B.  I would highly suggest you test that site recovery with a virtual manager before you actually run it live as it will save you a lot of downtime if you aren't learning how to recover it when the system is down for real.  The critical thing for you is not really the virtual manager as it is the multi-site bandwidth capacity compaired to your storage capacity.  If this isn't sufficient, the only solution is remote snapshots or some other solution on top that would compress the data so you don't actually have as many physical writes.