HA with *NO* common subnet between nodes?

[Automaton]

Is this possible? All the check boxes for heartbeat/sync are greyed out when I get to that step in the HA device setup.

Here is what I have...

Server1:
Amazon AWS availability zone "ABC"
NIC1 = 10.0.1.100 (primary)
NIC2 = 10.0.12.100 (want to use for heartbeat/sync)

Server2:
Amazon AWS availability zone ""XYZ"
NIC1 = 10.0.11.100 (primary)
NIC2 = 10.0.13.100 (want to use for heartbeat/sync)

Does the heartbeat/sync connection HAVE to be on a common subnet (For example, 10.0.12.xxx on BOTH servers)? Each server is in a different AWS availability zone. In Amazon AWS, I don't think I can have a subnet span across different availability zones.

[Anatoly (staff)]

Hi,

Yes, they should, otherwise they`ll not be able to communicate.

[jtmroczek]

May I request this get added as a feature request? I can see at least one scenario where routing on the sync channel would be useful, off-site DR.

I am sure you could work around the issue by using a VPN tunnel between the hosts, but there would be a rather significant latency penalty.

~joe

[Anatoly (staff)]

Well, you have the point, but the DR replication is not an HA - those are really different scenarios. We will have DR replication in our next release, that should arive in 1-2 moths. Stay tuned

[jwordtmann]

Hi,

I am jumping on this thread because I am in a similar predicament. Currently, I am using the latest build (Free version) from the StarWind website. Right now I have a PoC environment running in Amazon, using a VPC along with my instances being in different Availability Zones. This is a requirement for my PoC and I am trying to demonstrate the abilities of the iSCSI SAN product in hopes of convincing my superiors to purchase a license for the full version since EBS doesn't support iSCSI targeting on its own. Right now this is more or less a personal experiment so that I can bring something useful to the table when our business looks at moving the remainder of our products to the cloud.

That said, like OP I am stuck while creating my sync/heartbeat interfaces for HA configuration due to the interfaces needing to share the same subnet.

I am wondering if there is anyone on this board that has run into the same problem and if there is a workaround in the interim that I could employ. I wasn't quite clear if the inclusion of DR Replication in the next release will address this problem of mine, but any other solution that would work is fine by me. I would really appreciate it if someone could take the time to shed some insight to me as I am literally stuck at this point of my configuration.

Thank you!

[anton (staff)]

Please draw a detailed interconnect diagram (with all the subnets listed) say in Visio so we could try to think what could be done here. Thanks!

[jwordtmann]

Hello again,

I have actually discovered a workaround that allows for contiguous subnets across availability zones without using a Citrix bridge or any further configuration. The workaround involves creating a few custom subnets for the heartbeat and synchronization channels, and further involves changing the subnet masks around so Starwind iSCSI SAN is fooled into thinking they're on the same subnet. This is currently the configuration I am using and I will say that it is working well... Here's what I did:

1. Create a VPC in EC2
VPC Config
10.0.0.0/16 (255.255.0.0)

Just use the wizard here... Use the default layout, 1 public subnet.

2. Create some subnets for the primary, secondary, heartbeat and sync channels
Subnet Config
10.1.1.0/24 (255.255.255.0) - primary zone located in east1a... I use this as the subnet my Master iSCSI server lives in
10.1.2.0/24 (255.255.255.0) - secondary zone located in east1b... I use this as the subnet my Secondary iSCSI server lives in
10.1.4.0/28 (255.255.255.240) - primary zone located in east1a... I use this as the subnet for heartbeat/sync for my Master iSCSI server
10.1.4.16/28 (255.255.255.240) - secondary zone located in east1b... I use this as the subnet for heartbeat/sync for my Secondary iSCSI server
10.1.5.0/28 (255.255.255.240) - primary zone located in east1a... I use this as the subnet for dedicated heartbeat for my Master iSCSI server
10.1.5.16/28 (255.255.255.240) - secondary zone located in east1b... I use this as the subnet for dedicated heartbeat for my Secondary iSCSI server

MAKE SURE YOU ADD THESE SUBNETS TO THE ROUTE TABLE!!!

3. Assign the appropriate interfaces to your two Starwind servers
IP Config

Master
Primary NIC: 10.1.1.10 (This can be assigned via DHCP, if you configured that in your instance creation... really doesn't matter.)

Secondary Heartbeat/Sync NIC: 10.1.4.14 / 255.255.255.0 / gw 10.1.4.1 (This needs to be statically assigned so you can mangle the subnet mask... This is what fools Starwind into thinking the interfaces are on a contiguous subnet.)

Secondary dedicated Heartbeat NIC: 10.1.5.14 / 255.255.255.0 / gq 10.1.5.1 (This also needs to be statically assigned for the same reason as mentioned above for the Secondary Heartbeat/Sync interface).

PING ALL THE INTERFACES!!!

Secondary
Primary NIC: 10.1.2.10 (This can be assigned via DHCP, if you configured that in your instance creation... really doesn't matter.)

Secondary Heartbeat/Sync NIC: 10.1.4.30 / 255.255.255.0 / gw 10.1.4.17 (This needs to be statically assigned so you can mangle the subnet mask... This is what fools Starwind into thinking the interfaces are on a contiguous subnet.)

Secondary dedicated Heartbeat NIC: 10.1.5.30 / 255.255.255.0 / gq 10.1.5.17 (This also needs to be statically assigned for the same reason as mentioned above for the Secondary Heartbeat/Sync interface).

PING ALL THE INTERFACES!!!


From there you should be able to add the heartbeat and sync interfaces like I've done. It's not pretty, but I can confirm that this works entirely and I am currently employing this as a workaround until the multi-subnet functionality is supported in future Starwind releases. Basically, as far as Amazon is concerned, those subnets are indeed in different Availability Zones and are attached to their respective gateways. From the OS level, we have simply told Windows that the configured addresses are not /28 but actually /24, which then makes it seem like they are on the same subnet and thus allows Starwind to use them.

[Anatoly (staff)]

There is not such thing as "pretty" in this - "it works" -that is the main thing, so I`d like to say Thank You from the community since I`m sure that your way will be interesting for people.

[sharapov]

Well, you have the point, but the DR replication is not an HA - those are really different scenarios. We will have DR replication in our next release, that should arive in 1-2 moths. Stay tuned

Has DR replication been released? I just downloaded latest build and trying to setup replication between Primary and DR site and all of the check boxes for IPs when setting up "Interfaces for Synchronization and Heartbeat Channels" are greyed out.

Primary and DR sites are using different subnets and I can't use work around proposed to extend the same subnet across WAN.

[Anatoly (staff)]

Could you please drop me the full screenshot of the corresponding window?

[sharapov]

Could you please drop me the full screenshot of the corresponding window?

Please see screenshot below.

2014-08-22_10-46-58.jpg (79.45 KiB) Viewed 7846 times

[Anatoly (staff)]

Heartbeat/sync channels between nodes must be sharing the same subnet. DR replication has not been released yet as the final version - we will include some improvements.

I hope that makes sense.

[sharapov]

Hi Anatoly,

Yes, it does make sense, but unfortunately I do need it to be able to replicate between different subnets. I was told today that DR replication feature will be available at the end of September 2014. I'll test it once it comes out.

By the way do heartbeat and sync channels will need to be set on different physical network ports or both can be set on the same network port? I only have 1 available network port left on each of my servers in remote sites.

Thanks

[Anatoly (staff)]

Great question! I will talk to our R&D and I think we will improve this upon your request.

[barrysmoke]

any update on the DR?
I tried to create an async target on a cloud server, and I'm getting an error. I do see the free select option, and select the right interfaces on each host for sync, and heartbeat...but of course I don't have the ability to select a seperate heartbeat channel, since the cloud server has a single ip available, and the starwind node on my lan only has one nic with access to the internet.