ip configuration for 2node converged cluster

[qwertz]

Hi there!
I've got a question about where to use which IP address, and what it might affect. (sounds very simple, but i haven't understood it right how starwind handles it)
I'm rebuilding my home lab, consisting of 2 nodes(with starwind and Hyper-v on them) and I'm a little confused.

Here are my Networks:
- 10Gbit/s Sync / Heatbeat - 172.17.1.0 - Point 2 Point
- 10Gbit/s Migration(Hyper-V) / Heartbeat - 172.17.2.0 - Point 2 Point
- ~2Gbit/s 192.168.1.0 - Client Network / Heartbeat (LACP-Trunk with a VLAN-Trunk, Client network is one VLAN, IP addresses from here are listed in DNS)
- couple of other vlans that don't play a role here

what i'd like to get / build is that vms running on node a access the storage within node a, only changes get synced.
on node a crash, i'd expect the failovercluster booting my machines on node b, accessing the storage on node b.
IF the attached storage of node a fails, vms should stay on node a and use storage from node b accessing it over one of the 10gbit links.
and only looking at those 2 machines, i don't like to access any storage using 1gbit links.
(maybe i'll create an additional vlan for iscsi later, if i want to connect other servers to that storage, but not needed now)

Now... I'm unsure which IP addresses to use when:

a) adding them to the management console (does this affect anything?e.g. like the address that gets used if i click on "connect storage on the servers")
- if i use one of the addresses of the 10Gbits there will be problems if the partner node is down, as the local port also becomes down (point 2 point), and therefore the 172.17. network is unreachable)
- if i use the client network to add them(192.168.1.0)
- if i use loopback ( the local server is already listed as 127.0.0.1...)

b) when creating the cluster via "create new cluster"
- is it even a good idea to have node a and b listed under servers, and also under servers residing under Clusters.

c) adding them to the iscsi initiator
- should i use loopback and 172. something? on node a and b?

I would be very glad if someone could explain this to me, as i'm a bit unsure what i'm doing here...(and honestly: I don't like to try and error this)
Kind regards!

[Anatoly (staff)]

Hi,thank you for your questions!
Below are basic principles of configuring the HA storage cluster:
1. IP addresses and subnetworks should be configured taking into account sync channel total bandwidth: it should be not less than the sum of all throughputs of the client connection data links (in your case – 2 x 2Gb = 4Gb, so 10Gb sync channel is more than enough);
2. Your scenario is possible if you enable CSV;
3. Add nodes’ IPs from client network to StarWind management console (in your case 192.168.1.x). This IP address should be the result of normal ping of your node name.
4. Yes, it is a good idea to have both nodes listed in SW Mgmt console from ease of management stand point;
5. iSCSI initiator is a bit more difficult – the following setup is applied:
under discover portals you specify loopback first – 127.0.0.1 and set initiator as ‘default’. Then on both nodes set client network (iscsi storage sharing) communication. E.g., 127.0.0.1 > default + 192.168.1.1 > 192.168.1.2. Then set up target connections for each node storage using the following scheme:
- 127.0.0.1 -> default
- Client network ip -> the same client network ip (192.168.1.1 -> 192.168.1.1)
- Client network ip -> partner node ip (192.168.1.1 -> 192.168.1.2)
Repeat the same on partner node. As a result all of the targets should be listed as “connected” in MS initiator on both nodes.

I hope that answered the questions. If not - I`m always opened for the new questions.

[qwertz]

Hi there,
thanks for your time answering my questions, I'm very happy about that!
I've got a follow up question about point 5:
I thought it may be possible to use one of the 10gbit connections (point 2 point) using the setup i've described with the drawing.
Currently I'm using the Sync Network for that, if this is a very bad idea I could provide a third p2p connection.

Kind regards

[qwertz]

Hi there,
i've made a couple of additional screenshots to make clear what i'm trying.
Anatoly, I've mailed you the logfiles for this configuration(together with the cluster validation test results) from both clusternodes two days ago, did you recieve them?
(be aware, the Subnets differ as I don't post real world IPs in public forums.)

Kind regards

[qwertz]

, I'm sorry.
I got confused with the nicknames. Alex got the mail with the logfiles, should I send them to you also?

[qwertz]

*push*
I'm trying to get this "solution" running since 7 days. If there is any obvious thing I did wrong, it would be great if someone could tell me.
I sent you the logs via e-mail and didn't get any response to that so far.
I'm wondering if its a complex thing, or if you are too busy or are you missing any information?
Would be so great...

[qwertz]

Hi there,
I'm not sure if it helps, but:
If I create a non-ha storage on one of the two nodes the cluster validation test performs great.
So i think we can narrow it down to mpio / ha-storage.
Kind regards!

[Anatoly (staff)]

Let`s schedule the remote session to your system and verify everything.
Please send a quick email to support@ and we`ll schedule the exact time.
Please don`t forget to include the reference to this topic in the body of the email.