We do not use any raid in front of the NVMe drives, because of possible performance bottlenecks. We are using LVMThin on top of the drives.
-----Ursprüngliche Nachricht----- Von: David Sekne <david.se...@webtasy.com> Gesendet: Freitag, 19. April 2024 11:04 An: users@cloudstack.apache.org Betreff: Re: AW: Storage solution for Cloudstack Hi Swen, Since you need a pool to present to linstor on each node do you use any RAID (SW or HW) for the NVMe's (or you use ZFS / VG's)? Regards, David On 19. 04. 24 08:49, m...@swen.io wrote: > Hi David, > > 1. Are you running linstor just as HCI or you also have a dis-aggregated > cluster somewhere (and some recomendations regarding it)? >> we run only HCI clusters, so I cannot tell you a lot about standalone >> clusters. > 2. Whats the resource usage of linstor (even better if you have data for a > dis-aggregated cluster)? Info I got from linstor guys is not much (probably > true just for the replication) but looking at the blog from Andrei it seem to > be CPU intensive once you start utilizing it (RAM not so much). >> much less than what ceph is using on CPU and RAM. But of course you will see >> some impact on server resources like CPU and RAM. Reason is drbd and network >> stack. > 3. What kind of hardware are you running linstor on (CPU, Memory, disks)? >> we are using different cpus. Even some older like Intel E5-2680v4 and NVMe >> only. > Regards, > Swen > > -----Ursprüngliche Nachricht----- > Von: David Sekne <david.se...@webtasy.com> > Gesendet: Freitag, 19. April 2024 08:42 > An: users@cloudstack.apache.org > Cc: m...@swen.io; bryantian...@hotmail.com > Betreff: Re: Storage solution for Cloudstack > > Hello Swen, Bryan, > > We are just evaluating linstor at the moment as well (looks very promising). > I would like to run a standalone cluster (not HCI) just for linbit (so we > would use only diskless + nfs for secondary storage). > > I would have a couple of questions for both: > > 1. Are you running linstor just as HCI or you also have a dis-aggregated > cluster somewhere (and some recomendations regarding it)? > > 2. Whats the resource usage of linstor (even better if you have data for a > dis-aggregated cluster)? Info I got from linstor guys is not much (probably > true just for the replication) but looking at the blog from Andrei it seem to > be CPU intensive once you start utilizing it (RAM not so much). > > 3. What kind of hardware are you running linstor on (CPU, Memory, disks)? > > Thanks for your answers. > > Regards, > > On 30. 03. 24 00:56, m...@swen.io wrote: >> Hi Alexandru, >> >> We are also using Linstor on KVM hosts in Cloudstack and are very happy with >> it. >> >> Regards, >> Swen >> >> -----Ursprüngliche Nachricht----- >> Von: Alexandru Stan <alexandru.s...@lifeincloud.com.INVALID> >> Gesendet: Freitag, 29. März 2024 15:20 >> An: users@cloudstack.apache.org >> Betreff: RE: Storage solution for Cloudstack >> >> Hi Bryan, >> >> Interesting, thank you for the suggestion. Is this similar to ceph, >> installation wise? I mean a controller and storage nodes added to it? >> >> >> >> -----Original Message----- >> From: Bryan Tiang <bryantian...@hotmail.com> >> Sent: Friday, March 29, 2024 4:11 PM >> To: users@cloudstack.apache.org; users@cloudstack.apache.org >> Subject: Re: Storage solution for Cloudstack >> >> Hey Alexandru, >> >> We are using Cloudstack + Linstor right now and are live with a few clients. >> >> Its open source, performance is pretty great, and they support things like >> VM Snapshots, Volume Snapshots, Storage Replication, Encryption in >> Rest/Transit, VM HA and more. >> >> I think it suits what you’re looking for. >> >> Regards, >> Bryan >> On 29 Mar 2024 at 10:04 PM +0800, Alexandru Stan >> <alexandru.s...@lifeincloud.com.INVALID>, wrote: >>> Hi everyone, >>> >>> We have a specific scenario in witch we want to have a functional >>> Cloudstack cluster but we can't seem to find/decide on a working storage >>> solution, so if anyone can recommend a setup that meets all our >>> requirements, please do. So, we are using KVM for hypervisors and we want >>> to be able to offer vm snapshots functionality to the clients and also vm >>> HA - that excludes Ceph. If we go for NFS, we don't have redundancy, since >>> we're using nvme disks only and I believe we all know what ar the choices >>> of raid controllers for this right now. >>> We're considering Storpool too, but we've yet to receive a price list, and >>> we assume it's not cheap but who knows. >>> >>> Thank you! > -- > David Sekne > Systems Administrator > -- > DHH Infrastructure team > Webtasy, d.o.o., Slovenia > Plus Hosting Grupa d.o.o., Croatia > mCloud d.o.o., Serbia > > > -- David Sekne Systems Administrator -- DHH Infrastructure team Webtasy, d.o.o., Slovenia Plus Hosting Grupa d.o.o., Croatia mCloud d.o.o., Serbia
