| 10 Mar 2024 |
 adamcstephens | i don't have shared storage so i haven't explored it myself | 18:40:42 |
 hexa | Yeah, but lvm means having shared central storage | 18:49:55 |
| hexa | Which blows | 18:49:59 |
| adamcstephens | don't you already run ceph? | 19:00:53 |
| hexa | I do | 19:02:58 |
| hexa | was just trying to understand the alternatives | 19:03:12 |
| adamcstephens | https://github.com/lxc/incus/issues/564 | 19:03:48 |
| adamcstephens | https://github.com/lxc/incus/issues/344 | 19:04:05 |
| hexa | glusterfs is some weird distributed filesystem 😄 | 19:05:09 |
| hexa | linstor is iirc the firm behind drbd | 19:05:19 |
| adamcstephens | yep | 19:05:26 |
| hexa | or linbit is | 19:05:31 |
| hexa | I don't want to go back to drbd 😄 | 19:05:45 |
| hexa | so what is linstor | 19:05:53 |
| hexa |
LINSTOR has multiple uses, but it is typically used to manage a large number of DRBD replicated storage resources
| 19:06:53 |
| adamcstephens | i've never used drbd 🤷♂️ | 19:07:16 |
| hexa | way back, when we had two node hypervisor clusters over 10y ago | 19:07:30 |
| adamcstephens | yeah, gluster is also an old filesystem. but it's dying | 19:10:00 |
| adamcstephens | i think the real problem is clustered/distributed filesystems are hard and RH is probably the only one making any serious money with them | 19:12:24 |
| adamcstephens | (in the open source space) | 19:12:40 |
| hexa | so what clusterlvm does is make sure that in a clustered lvm setup that only one node can write metadata | 19:16:09 |
| hexa | they share the same block device, which can be attached by any means, but most likely iscsi | 19:16:27 |
| hexa | that only makes sense if you have a chonky multihead storage cluster | 19:16:41 |
| hexa | I'd rather do hyperconverged setups, and that is where ceph shines | 19:17:06 |
| hexa | the main benefit is that I can rely on commodity hardware | 19:18:57 |
| hexa | anyway, one weird thing I imagined is zfs send/recv style migrations | 19:19:26 |
| hexa | I think proxmox could do that | 19:19:36 |
| hexa | and it might be appealing since zfs has alot less moving parts than ceph | 19:19:47 |
| adamcstephens | at least for non-clustered incus, i think it does use zfs send/recv for migrations | 19:34:21 |
| adamcstephens | it may for clustered too when it's not a shared volume | 19:34:42 |
