Power Outage Considerations and ZFS Support

Software-based VM-centric and flash-friendly VM storage + free version

Moderators: anton (staff), art (staff), Max (staff), Anatoly (staff)

Post Reply
User avatar
NewbAdmin
Posts: 3
Joined: Sat Jan 04, 2025 9:37 pm

Sun Jan 05, 2025 11:40 pm

Hello, StarWind community. I have been working on a proof of concept to get my home environment off of TrueNAS and onto an HA storage backend. StarWind seems like the best (and only) free solution that suits my needs. Unfortunately (for me anyway) there are some potential deal breakers. I'm hoping someone can help me out with some of it.

Here are the items that I need some guidance with:
Is there a functional difference between Virtual SAN and the virtual HCI appliance once setup has been completed? They both look the same in the quickstart guides, but I find myself wondering if some of my other issues (see below) might be addressed by changing products.

Is there a way to conduct a graceful/gentle shutdown of the VSAN cluster during a power outage? I can get a UPS if I need one (which seems likely), but I do not have access to a backup power source for sustained power loss. The closest solution I could find was some documentation on integrating with APC PowerChute, but it's from 2018 and appears to be for VSAN deployed as an application on Windows which (I think) has been deprecated.

Can VSAN be configured to only replicate missing/differing information to unsynchronized nodes? If I simulate a power outage by abruptly shutting down the cluster then starting it back up, I need to mark one node as synchronized and then wait for the partner node to be completely overwritten. This leaves the cluster in a vulnerable state for a longer period of time, so I would like to avoid it if at all possible. To simplify this scenario, assume that all hosts connecting to the cluster are using MPIO in 'Fail Over Only' mode instead of 'Round Robin' or 'Least Queue Depth' modes.

As I mentioned above, I'm coming from TrueNAS which has very good support for a lot of advanced ZFS features. VSAN doesn't seem to have nearly the same level of robust support. In fact, it was very challenging to even get vdevs added to a pool. I tested by creating a 3-disk RAIDZ1 pool. I was able to add another 3-disk vdev to the pool by going to 'Storage pools/Manage pool/Layout' in the web UI and changing the pool layout, but I can only do that once. After that, the 'adjust pool layout' button is disabled. How to add disks to the pool from that point on is incredibly difficult to decipher and I have only been able to add hot spares. Is this a bug in the current version of the CVM, or is ZFS support just not very good? I was able to manually add a vdev via the command line, but this is very undesirable since I don't know what I could be potentially breaking by doing so.

I know that's a lot to ask, or at least it seems like it is, but between the various different iterations of VSAN over the years and a lack of updated documentation for similar features in newer versions, I'm at a loss here. The power loss issues in particular are real show stoppers for me since this is a home environment and we get some pretty ugly winter storms that can knock power out for a few hours at a time. I appreciate any advice and guidance anyone is willing to offer.
yaroslav (staff)
Staff
Posts: 3424
Joined: Mon Nov 18, 2019 11:11 am

Mon Jan 06, 2025 4:39 am

Welcome to StarWind Forum.
Is there a functional difference between Virtual SAN and the virtual HCI appliance once setup has been completed?
The short answer will be support. StarWind vHCA comes bundled with ProActive Support https://www.starwindsoftware.com/starwi ... ve-support. Also, there's broader assistance during implementation (e.g., spacing and migration assistance).
Is there a way to conduct a graceful/gentle shutdown of the VSAN cluster during a power outage?
See https://knowledgebase.starwindsoftware. ... production. You can also use that APC's software https://www.starwindsoftware.com/resour ... owerchute/.
Can VSAN be configured to only replicate missing/differing information to unsynchronized nodes?
There are two times of StarWind synchronization full and fast. Fast is performed when one node is down. Full is performed when there is a power outage. You can mitigate the risk of full synchronization by sacrificing part of your performance and using a continuous disk journal (https://knowledgebase.starwindsoftware. ... a-devices/).
If I simulate a power outage by abruptly shutting down the cluster then starting it back up, I need to mark one node as synchronized and then wait for the partner node to be completely overwritten.
This is the most extreme scenario and does not really happen that much unless you have write-back cache configured . If you do mark the node as synchronized, the only way out of it will be full synchronization by design.
This leaves the cluster in a vulnerable state for a longer period of time
You can mitigate the impact of full synchronization by using smaller devices: more chances that more data will be HA within the same amount of time.
To simplify this scenario, assume that all hosts connecting to the cluster are using MPIO in 'Fail Over Only' mode instead of 'Round Robin' or 'Least Queue Depth' modes.
MPIO has nothing to do with it. Non-synchronized devices are not available over iSCSI. Leave least queue depth: the replication partner will be available over iSCSI when replication is over.
VSAN doesn't seem to have nearly the same level of robust support.
It does have ZFS support. You can try using CLI.
I will discuss UI limitations with my colleagues.

As of the documentation, we are working on it. You are welcome to request documentation by filling in this form https://www.starwindsoftware.com/support-form

Good luck with your project.
User avatar
NewbAdmin
Posts: 3
Joined: Sat Jan 04, 2025 9:37 pm

Sat Jan 18, 2025 4:58 am

I have been able to mitigate the power outage issues by adjusting my configuration in accordance with your recommendations. I have also reviewed the documentation for the Windows-based management console (and PowerShell help topics) which I had initially misunderstood to be deprecated and no longer useful. Of course, with that said, it does appear that StarWind is a little between products at the moment.

I would like to get some clarification on these points:
- Is the Windows management console being phased out in favor of the Web UI, or are they meant to complement each other? I ask because there is a stark difference in features between them.
- Based on the webinars, the now legacy SAN & NAS product seems to have better ZFS configuration support. Is the VSAN web UI expected to reach feature parity with SAN & NAS?

Let me know if I should start a new topic for these additional questions. I would also be willing to provide additional information/screenshots of the ZFS limitations in VSAN if it is helpful.

Thanks
yaroslav (staff)
Staff
Posts: 3424
Joined: Mon Nov 18, 2019 11:11 am

Mon Jan 20, 2025 6:44 am

Hi,

Both Windows-native and CVM are supported.
- Is the Windows management console being phased out in favor of the Web UI, or are they meant to complement each other? I ask because there is a stark difference in features between them.
They are from different products originaly. Windows-based console now has more features; they should gradually appear in StarWind CVM web console.
- Based on the webinars, the now legacy SAN & NAS product seems to have better ZFS configuration support. Is the VSAN web UI expected to reach feature parity with SAN & NAS?
StarWind SAN and NAS is legacy product.
ZFS can be enabled in the Ubuntu-based CVM.
Let me know if I should start a new topic for these additional questions. I would also be willing to provide additional information/screenshots of the ZFS limitations in VSAN if it is helpful.
You can stick with this topic.
Thanks for your feedback!
Post Reply