Hyper-V Hyper-Converged Setup Not Synchronizing

seanp · Thu Aug 04, 2022 11:21 pm

Hello! After a few days and a steep leaning curve, I have what I assume is a functional system. Except that the mounted HA disk will not sync unless I offline the disk and then online the disk again.

3 nodes, each has 2x NVMe (soft-raid) for the OS and 4x SATA SSD. The servers have teamed 1G NICs for the OS/VM network, and teamed 10G NICs for storage replication. The 4 disks are pooled/vdisk'd 9simple), and volume'd. They are all mounted in C:\Storage. I used the CreateHA_3.ps1 script to create a 3 node HA cluster (I assume). I added all of the iSCSI targets to the initiators, and the online'd/initialized the disk. It is mounted in C:\VMStorage.

I can create a text file in the C:\VMStorage folder on one node, and it will not replicate on its own. Instead, I have to go to another node into disk management, offline the HA disk, then online it again. Then the text file replicates. This appears to be the same with virtual machine disks, as I created a VM on the first node and nothing would replicate until I did the manual offline/online on another node.

Is there something I am missing here? A scheduled task or something I need to set up to get them to replicate on their own? Was considering just trying to script it every hour or something just to have some replication in place, but if I'm missing something built-in (I think I am) I'd rather have that.

Thanks!

Fri Aug 05, 2022 5:12 pm

Welcome to StarWind Forum!

3 nodes, each has 2x NVMe (soft-raid)

What software RAID do you use?

and teamed 10G NICs for storage replication

Teaming for iSCSI and sync is against StarWind's best practices. I would always recommend going with Multipathing instead of LACP or any other kind of link aggregation if we are talking about iSCSI networks. Storage networks require minimum latency, but different link aggregation implementations can act under different logic sets leading to inconsistent results in stability and performance.
Basically, LACP transfers frames through the separate physical connections and ensures the correct frame order by comparing the hash of each transfer. Keep-alive frames are transferred time to time in order to ensure the physical network availability. Such implementation does the job in terms of redundancy, but, in most cases it loses in performance in comparison to Multipathing due to increased latency.
See more at https://www.starwindsoftware.com/best-p ... practices/.

I used the CreateHA_3.ps1 script to create a 3 node HA cluster (I assume)

Can I see a script?

I can create a text file in the C:\VMStorage folder on one node, and it will not replicate on its own.

Unlike file-level protocols used in NAS, iSCSI is a block-level protocol and it cannot arbiter read-write access to an iSCSI device connected to multiple servers. In order to provide the access to one device from multiple servers, the device needs to have a clustered file system. While VMFS is a clustered file system and no additional actions should be done to see updated data on datastore, iSCSI storage, connected to the nodes in a Microsoft failover cluster should be managed by the cluster and formatted as CSVFS if used as Cluster Shared Volume (CSV). Such an approach allows to share single storage device between the nodes in the cluster and get updated data on them.
https://www.starwindsoftware.com/starwind-faq

seanp · Fri Aug 05, 2022 5:50 pm

Thank you for your reply. Here is the script I am using: https://gist.github.com/sean-perryman/7 ... ed35034c37

I am just using Windows mirroring for the OS "raid". The server has a software raid on the motherboard, but I've disabled that (it was passing RAID as the bus type, now it passes SATA as the bus type).

As far as NIC teaming goes, they are set as "switch independent", and basically just acting as a backup from what I understand. Not doubling bandwidth or anything. Just wanted a little redundancy. Should I drop the team and just configure the 2nd NIC with it's own IP (e.g. 172.16.32.1 and 172.16.32.11 for server 1)?

I guess I'm not following you here. I have approximately 7T of usable space on each server. When I ran the script above and connected everything to each-others' iSCSI targets, I was presented with 1 disk in disk management. I assumed this was the ha/replicated/redundant disk created by the system. Is that not correct?

I came across a tutorial about setting this up and it mentioned enabling Cluster Shared Volumes, but that doesn't appear to be an option on this setup (perhaps it was already enabled?). I could certainly tear it all down and start over if that is your suggestion, I just need to get something working. I have all their VMs crammed on a single VMware server at the moment and I'd hate to have something happen before I could migrate them into the new environment.

Mon Aug 08, 2022 5:06 pm

Thanks for your update.
To improve the script, please consider using a redundant IP as a heartbeat, also make sure your setup aligns with https://www.starwindsoftware.com/system-requirements.
On top of that, do not keep the StarWind *.imgs on C:\.

I am just using Windows mirroring for the OS "raid".

This is not the best option. Please consider using Storage Spaces instead. Consider creating a hardware RAID if possible.

Not doubling bandwidth or anything. Just wanted a little redundancy. Should I drop the team and just configure the 2nd NIC with it's own IP (e.g. 172.16.32.1 and 172.16.32.11 for server 1)?

Yes, please use different links instead of teaming them.

Is that not correct?

This is correct, but that's not how NTFS works. Please see a similar question at https://www.starwindsoftware.com/starwind-faq.

I came across a tutorial about setting this up and it mentioned enabling Cluster Shared Volumes, but that doesn't appear to be an option on this setup (perhaps it was already enabled?)

Did you install the Failover Clustering feature? Could you please share the role for this server? There could be a ready-to-go guide for your use case.

To get a better understanding of StarWind VSAN operation, I'd suggest having a call with one of our techs and setting up a demo. Having such sort of a tech call will be really beneficial if this is planned for a production system. Please drop us an email at support@starwind.com, using this thread and 746623 as your references.