This message was deleted.
# harvester
a
This message was deleted.
b
It is in fact 2024 - that is a true statement.
πŸ‘€ 1
r
Wouldn't make much sense if it wasn't πŸ˜‰
s
Hi @red-stone-23003, we currently support non-root disks from 3rd-party storage like Rook. Ref: https://harvesterhci.io/kb/use_rook_ceph_external_storage
r
Ooooh, really? I've been following Harvester for a while, and lack of support for Ceph was one of the reasons I wasn't using it
Where is NFS on the roadmap? It's the primary reason we aren't deploying it in production on my team - we just deployed a bunch of NetApps, but can't use it with Harvester
m
@red-stone-23003 NFS isn't on the roadmap right now (specifically) but integration for third-party storage for root disk is. In the mean time: https://harvesterhci.io/kb/install_netapp_trident_csi You can explore those NetApps you've got
r
We looked at Trident w/Harvester the other day, but the note about having to use Longhorn the boot volume made us decide Harvester wasn't the right choice for us
m
I'm sorry to hear that, it's just the LH requirement that made you think otherwise?
r
Yes - we don't want any of the VMs stored locally, so the requirement is 100% usage of external storage (not counting the hypervisor root install of course) - whether that's NFS, or some Netapp specific storage like Trident
to be clear, it's the documentation here: https://harvesterhci.io/kb/install_netapp_trident_csi/
3rd party storage classes (including those based on Trident) can only be used for non-boot volumes of Harvester VMs.
should it be possible to use Trident for the boot volumes of VMs as well, we might be interested in using Harvester
m
Apologies, just on a call πŸ˜„ https://github.com/harvester/harvester/issues/1199
r
haha, no worries. Same here πŸ˜›
s
Hi @red-stone-23003, would you mind opening an enhancement GH issue for the 3rd-party root support? I saw a similar response on https://github.com/harvester/harvester/issues/1199#issuecomment-1249963051. But, AFAIK, seems we do not have a dedicated issue for it. Thanks!
r
You guys are providing a solution that almost everyone else on the market has abandoned - namely, a VM Hypervisor as an appliance. Our other options are: β€’ Microsoft (we need a linux solution) β€’ vmware (not in our budget) β€’ Proxmox (security issues, like passwordless root-access between all nodes) β€’ OKD (poorly documented, doesn't really work) and... that's it. All the other solutions we've looked at are either paid, or abandoned in favor of containers. Harvester can have a niche, but the lack of support third-part or other storage solutions is limiting.
m
@red-stone-23003 thanks for the feedback. Would you be open to submitting an issue to discuss with the engineering team + myself there?
r
Hi! I'll try to get around to it πŸ™‚
m
That'd be awesome πŸ˜„ (although I'm just going to double check we don't have anything already there that you can add weight to)
❀️ 1
Righto: https://github.com/harvester/harvester/issues/4688 this is somewhat relevant to your request, but there's a few coming up regarding NFS. Shuffling through them now...
Just so I understand properly, you just don't want to store any VMs locally. That's the entire use case here?
r
Right. NFS is kind of the lowest common denominator IMO. Though if there's plans to support root volumes on Trident, that might work for us too
m
I don't see why that wouldn't be the case with the root volume work πŸ€”
r
hmm?
according to the previously linked docs, harvester only supports Netapp Trident for *no*n boot volumes
r
just to be clear, that includes supporting the boot volumes on other servers, right? Not like "we support mounting iSCSI volumes to harvester and then running longhorn on them like they were local" - because that's a little different
m
What do you mean by "boot volumes on other servers"? Not sure I understand.
r
I mean like, where you mount remote storage - NFS, CIFS, CephFS, RBD, whatever, and that's where your VMs storage is, and not locally
> 3rd party storage classes (including those based on Trident) can only be used for non-boot volumes of Harvester VMs. Makes it sounds like you support mounting additional volumes to VMs based on Trident, but the root/boot volume of VMs have to be on local storage
m
So (I just went to double check before I said anything - Slack switching is fun) we would likely start by supporting iSCSI LUN. So yes, the server is diskless (has no hard drive) so the boot volume has to be on the other storage array/appliance.
r
no worries - I'm at work too, so I'm expecting this to be a bit asynchronous πŸ™‚ I appreciate your time!
πŸ™Œ 1
we have disks for the OS to install on, just not for any hyperconverged storage. As long as that setup works, we are golden, and I can try to put Harvester on the road-map to PoC when you release version 1.4.0 πŸ™‚
m
🫑 call me come October, pal!
r
will do πŸ™‚
m
The roadmap is always subject to change, but I hear the product manager is really cool and approachable
r
πŸ˜›
m
Aw you're supposed to fall for that.
r
Oh, I did! I totally believe you!
are you the project manager? lol
m
I am the Product Manager 🀣
I just am brazen enough to call myself cool and approachable on a public forum
r
hahaha. I'll give a +1 to that yelp review then πŸ˜†
m
✊ job well done, team. Everyone take the day off.
r
Haha, I tell people all the time about how Intel has good-looking engineers πŸ˜›
(Like that commercial from the IT crowd "Now with better looking drivers!" )
🀣 1
m
Who's from Intel!?
r
Me!
m
Now you've outed yourself!
r
darn! Foiled again!
m
Intel wants third-party storage for Harvester hey, well well well
How the turns have tabled
πŸ˜† 1
r
Haha, well don't quote me - I can speak for like, one small team, not all of Intel πŸ‘€
did someone else from Intel tell you something different? Lemme at 'em!
m
Damn, so, I should probably delete the slide deck called "Layne on Slack Says"
r
that depends on how far along my offer letter is πŸ˜‰
m
You're going to get ONE BILLION DOLLARS and start soon.
r
Oh dang! Make sure you email me that in writing. I could finally afford a tiny home!
m
The millennial dream
Quick question, for the really cool and approachable PMs knowledge bank, what "small team" in Intel would that be? It'd be great to know who is using it and in what kind of capacity
πŸ˜› 1
(or thinking about POCing it)
r
sorry, headed home. Back online now
Datacenter AI, System Design & Enablement. And I mean small - there's like, 4 people on my team total
but I hear other teams are using it - one reason we are looking at it πŸ™‚
m
Interesting! Well, if you want to talk more about it: slack or email door open. Would love to hear about what your needs are and whether or not we can actually solve some problems here.
r
Thank you!