Best posts made by TheDeepStorage

@scottalanmiller said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

So two nodes is the small end for a hyperconverged appliance (or one that you build yourself with hypervisor + Starwind VSAN), I think I'm clear there. But what about upper limits? A lot of hyperconverged appliances cap out around ten or twelve nodes in a single cluster. What are the limites from the Starwind side if I'm building Hyper-V or VMware clusters? Is it the same and a Starwind limitation? Do I just get to go to the limit of the hypervisor?

With VMware we can push it to the limit of 64 nodes using VVols. With Hyper-V, in theory, StarWind handles 64 nodes just fine, but the best practises of Microsoft suggest 1 CSV per node, which means 64 CSVs, which means one very sad IT admin with a twitching eye and chronic headaches.
The short version would be: from our side, the hypervisor is the limit.

That would split up failure domains pretty heavily, though. Impacts to one part of the cluster would not ripple through to other parts.

Definitely, if you can manage this cluster, it will be a very resilient environment.
Ultimately, we might consider a promotion of providing Xanax to any admin that configures 64 CSVs free of charge.

Thank you for all the questions, we've had a great time (and we hope you did as well!). Also, I will abuse the situation to advertise @ABykovskyi and his webinar about the Linux VSA that will be happening tomorrow: https://www.starwindsoftware.com/dedicated-live-demo-session

Hello guys! We are looking forward to all your questions and me, @LaMerk and @ABykovskyi will do our best to fill you in on where we stand currently, what we are looking at doing in the future and respond to any other questions you have.

@dafyre said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@coliver said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@DustinB3403 said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@dafyre said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Is the Two-Node setup for Starwind an Active/Active setup or active/passive?

StarWind's two-node setup is an active-active scenario.

With this setup is there any risk of a split brain scenario occurring? If so what protections are built in to ensure that the system normalizes it's self?

Instead of forcing people to use a 3rd host as withness we use a hearbeat channel (preferably set up on a separate physical NIC to avoid a SPOF) that constantly has the nodes pinging each other. This way if the synchronization channel fails the HA device that is second by priority will be disabled to avoid split-brain.

What if it fails due to the primary hosts failure? Sudden power outage etc. In a two node setup would both nodes then go down?

Both hosts are constantly in sync,and each VM has access to it's virtual disk on both hosts. If one of the node fails, then the cluster just continues working if it's VMware FT or the VMs failover and everything continues working in case of Hyper-V or VMware HA.

How would we configure the iSCSI connection in Hyper-V for this? Using MPIO?

Exactly. We actually have a lot of guidance to configuring MPIO properly on our webiste. Here is a Knowledge Base article with details: https://knowledgebase.starwindsoftware.com/guidance/how-to-configure-mpio-with-starwind-correctly/

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

So two nodes is the small end for a hyperconverged appliance (or one that you build yourself with hypervisor + Starwind VSAN), I think I'm clear there. But what about upper limits? A lot of hyperconverged appliances cap out around ten or twelve nodes in a single cluster. What are the limites from the Starwind side if I'm building Hyper-V or VMware clusters? Is it the same and a Starwind limitation? Do I just get to go to the limit of the hypervisor?

With VMware we can push it to the limit of 64 nodes using VVols. With Hyper-V, in theory, StarWind handles 64 nodes just fine, but the best practises of Microsoft suggest 1 CSV per node, which means 64 CSVs, which means one very sad IT admin with a twitching eye and chronic headaches.
The short version would be: from our side, the hypervisor is the limit.

@QuixoticJeremy said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Oh, thought of another. What connection protocols are supported by Starwind? Like iSCSI, I know, as everyone talks about it. Anything else like SMB?

Quite a few actually. We support ISCSI, SMB 3.0, NFS 4.1, VVols just to name the main ones.

Oh and by the way, we will be hanging out on Mangolassi even after the AMA, so if you have any late questions afterwards, feel free to tag us.

@scottalanmiller I knew it had to be fluffyctulhudeepstorage!

@DustinB3403 said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@dafyre said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Is the Two-Node setup for Starwind an Active/Active setup or active/passive?

StarWind's two-node setup is an active-active scenario.

With this setup is there any risk of a split brain scenario occurring? If so what protections are built in to ensure that the system normalizes it's self?

Instead of forcing people to use a 3rd host as withness we use a hearbeat channel (preferably set up on a separate physical NIC to avoid a SPOF) that constantly has the nodes pinging each other. This way if the synchronization channel fails the HA device that is second by priority will be disabled to avoid split-brain.

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

How does the caching work on Starwind storage? I've read that I can use RAM cache, and obviously there are the disks in RAID. Can I have an SSD tier between the two? Can I have multiple tiers like a huge RAID 6 of SATA drives, a smaller RAID 10 of SAS 10Ks, a smaller SSD RAID 5 array and then the RAM on top?

StarWind can utilize both L1 cache on RAM and L2 cache on SSDs.
In regards to a specific configuration, as an example: you can have a huge RAID 6 array for your coldest data, then a moderate RAID 10 10k SAS array for your day-to-day workloads, a small RAID 5 of SSDs for I/O hungry databases and then top it off with RAM caching. That being said we do not provide automated tiering between these arrays and you would assign everything to each tier specifically. You could easily use Storage Spaces 2016 with StarWind for that functionality. Just make sure not to use SS 2012, since the storage tiering functionality on it sucked was suboptimal and lead us to the decision of not doing automated tiering in the first place.

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

How does the caching work on Starwind storage? I've read that I can use RAM cache, and obviously there are the disks in RAID. Can I have an SSD tier between the two? Can I have multiple tiers like a huge RAID 6 of SATA drives, a smaller RAID 10 of SAS 10Ks, a smaller SSD RAID 5 array and then the RAM on top?

StarWind can utilize both L1 cache on RAM and L2 cache on SSDs.
In regards to a specific configuration, as an example: you can have a huge RAID 6 array for your coldest data, then a moderate RAID 10 10k SAS array for your day-to-day workloads, a small RAID 5 of SSDs for I/O hungry databases and then top it off with RAM caching. That being said we do not provide automated tiering between these arrays and you would assign everything to each tier specifically. You could easily use Storage Spaces 2016 with StarWind for that functionality. Just make sure not to use SS 2012, since the storage tiering functionality on it sucked was suboptimal and lead us to the decision of not doing automated tiering in the first place.

Okay so basically there are two cache layers, L1 RAM and L2 SSD Array, and then you would have to "manually tier" anything beneath that?

Any options for sending cold storage out to cloud like S3 or B2, which is popular here?

Yes, exactly. Or if you prefer automated tiering 2016 Storage Spaces play nicely with StarWind.

We have several options to offload data to cloud:
-Our own asynchronous replication, which allows to do block level replication to the cloud.
-We can provide a cloud gateway solution as part of our appliance infrastructure, which is plugged into the SATA bus and presents the cloud as a local cold storage tier.
-We have an offering with our partners from Veeam where you can offload data to the cloud using virtual emulations of physical tapes. This allows to kill 2 birds with 1 stone: you store backups (for example) in the cloud and have ransomware protection, because cryptolocker doesn't target physical tapes (for obvious reasons).

@Reid-Cooper said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

-Our own asynchronous replication, which allows to do block level replication to the cloud.

Is this what we would normally look at for an off-site DR solution? What would be the "go to" for that?

Let's say I have a two node Starwind HC setup at my office. And I want a DR site, would I do a single node at a colo or a second office and use the asynchronous replication feature to get the data to it? What does that solution look like?

It sounds, like you might start architecting StarWind deployments. That's exactly what we would suggest. Just add a node at a remote location and set it up as an asynchronous replication partner. Schedule the snapshots and your DR environment is good to go.

@hobbit666 said in Deploying a 2-node StarWind VSAN Free for VMware ESXi 6.5:

Do the nodes need to be identical?

According to our best practises-yes. Since the hosts are used in an active-active scenario, having one of the nodes slower than the other will have several significant consequences:
-You will have delays for all the operations. Reads are performed from both hosts and writes are done on both nodes as well, using round-robin. If one of the nodes is the bottleneck, you will see latency and a considerably lower I/O overall
-If one the faster host fails, all your environment will failover to the slower node and chances are high it will not be able to handle it

So in short, you CAN use non-identical nodes, but that doesn't mean you should.

@hobbit666 said in Deploying a 2-node StarWind VSAN Free for VMware ESXi 6.5:

Thanks just that I might have some kit becoming available but they are all different Dell models, just wondered if the Hard Drive space/RAID matched that would be OK for a test lab,

e.g.
R610, 64GB RAM with 4x 600GB SAS
R410, 64GB RAM with 4x 600GB SAS

Sounds like it will work just fine for a test lab, anyways, once you get down to testing, DM me and I'll get you in touch with an engineer, who will review your environment to make sure everything will work perfectly.

Looks like I'm a bit late to the party and Scott tagged anyone but me:( Just looking through the requirements, I see one of the typical StarWind use cases. Also, we're extremely user friendly both in terms of configuration and in terms of licensing.
So generally, a vendor offering an HA solution will hate you, for asking whether you need HA. We won't, we'll actually suggest to look at your data and determine how much actually needs to be stored in an HA storage pool. This will let you save soooo much on storage costs, both in terms of actual drives and in terms of StarWind licensing. Also, we have pre-built appliances which have ProActive support and are sized for your specific requirements. Overall, I think it's worth checking your options with us. I'll leave some links for reference:
https://www.starwindsoftware.com/starwind-virtual-san
https://www.starwindsoftware.com/starwind-hyperconverged-appliance

Hi Kelsey,

First of all, as previously mentioned in the thread, the IPOD scenario is absolutely atrocious no matter how you look at it. You pay more to get more hardware to manage to get lower resiliency. Here is a very short paper on why it's particularly important to have HA or FT in a virtual inrastructure:
https://www.starwindsoftware.com/fault-tolerance-and-high-availability-page

A way to play around IPOD is replicating local storage across the hypervisor hosts. Here is a doc describing how it's done:
https://www.starwindsoftware.com/whitepapers/starwind-virtual-san-whitepaper.pdf

Also, our forums are pretty helpful and you can ask different virtualization, storage, HA questions there:
https://forums.starwindsoftware.com

@dafyre said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Is the Two-Node setup for Starwind an Active/Active setup or active/passive?

StarWind's two-node setup is an active-active scenario.

RAID 5 for SSD is a yes. Space-efficient, yet performs decently, if you need the extra redundancy and make the system future-proof just get a second server with SSD in RAID 5 and balance the storage efficiency with replicating part of the data with a VSAN type solution. StarWind can make it easier from a cost perspective due to us having a free version, but ultimately any VSAN will do the trick.

@coliver said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@DustinB3403 said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@TheDeepStorage said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@dafyre said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Is the Two-Node setup for Starwind an Active/Active setup or active/passive?

StarWind's two-node setup is an active-active scenario.

With this setup is there any risk of a split brain scenario occurring? If so what protections are built in to ensure that the system normalizes it's self?

Instead of forcing people to use a 3rd host as withness we use a hearbeat channel (preferably set up on a separate physical NIC to avoid a SPOF) that constantly has the nodes pinging each other. This way if the synchronization channel fails the HA device that is second by priority will be disabled to avoid split-brain.

What if it fails due to the primary hosts failure? Sudden power outage etc. In a two node setup would both nodes then go down?

Both hosts are constantly in sync,and each VM has access to it's virtual disk on both hosts. If one of the node fails, then the cluster just continues working if it's VMware FT or the VMs failover and everything continues working in case of Hyper-V or VMware HA.

@ABykovskyi said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

@travisdh1 said in Starwind AMA Ask Me Anything April 26 10am - 12pm EST:

Apparently I'm going to be starting with the unusual questions. In your V2V Converter, you have the option to convert to Microsoft VHD, why the Microsoft VHD and not just VHD?

Thaks for great question. It should actually work with XEN as well.

Looks like we'll have to double check this and then make some changes to the V2V to reflect it supports both.