This message was deleted.
# harvestera
adamant-kite-43734
06/29/2023, 7:16 PMThis message was deleted.
b
better-garage-30620
06/29/2023, 8:59 PMi would be nervous using harvester in production without zero-downtime upgrades.
better-garage-30620
06/29/2023, 9:01 PMhttps://blog.stevecoinc.com/2023/05/harvester-hci.html has a summary of my thoughts, if you're interested.
b
blue-kitchen-51801
06/30/2023, 4:00 AMhey, creating certs using LetsEncrypt shouldn't be hard because Harvester comes with a built-in RKE cluster and you could deploy Cert-manager to generate the certificate using LetsEncrypt, ever though, it would be nice to have some sort of add-on to configure this out of the box
🙌 1
b
broad-knife-10346
06/30/2023, 12:46 PMThanks for sharing @better-garage-30620!
b
big-judge-33880
06/30/2023, 11:01 PMDidn’t have any workload downtime on our last 1.1.1 to 1.1.2 upgrade FWIW
👍 1
big-judge-33880
06/30/2023, 11:03 PMThere is a lot of very time consuming longhorn replica rebuilding going on during upgrades though, so it’s quite a lengthy process in a six-node cluster with quite a few large volumes, even on a nice 10Gbe separate storage network with jumbo frames
👍 1
big-judge-33880
06/30/2023, 11:05 PMReally looking forward to 1.2, particularly since longhorn has become a lot more reliable in the 1.4 branch
👍 1
p
plain-painter-84885
07/01/2023, 1:51 AMWe are running a 20 node harvester cluster on with 100Gbps nics, 1TB of memory, 5 nvm disks, and 128core per node we would love to offer our input.
b
broad-knife-10346
07/03/2023, 11:05 AMThanks for sharing @big-judge-33880 !
broad-knife-10346
07/03/2023, 11:06 AMI would really appreciate that, if you have time @plain-painter-84885 !
s
salmon-city-57654
07/03/2023, 4:31 PMHi @broad-knife-10346,
Like the above sharing, we should have done the zero downtime upgrade.
IIRC, upgrade from v1.1.1 to v1.1.2 should not have downtime time.
We are now working on v1.2.0. Surely we would like to offer the zero downtime upgrade.
Could you share more about what the workloads are running? (VMs or containers)
And for the SAN storage, would you like to add to the Longhorn as storage? or external storage then operated by CSI?
b
broad-knife-10346
07/04/2023, 11:58 AMThanks @salmon-city-57654 ! Sure!
The workloads are running VMs, and the VMs are again running containers, with docker swarm as orchestrator. The containers are read heavy and/or compute heavy. (We want to move over to kubernetes soon.)
This is a new site we are configuring, so the SAN doesn’t really exist. The proxmox HA cluster + TrueNAS is just our regular architecture. Therefor no need to migrate existing data. However, if we are happy with the setup and decide to change our other sites, then we want to transfer everything from the existing SANs to the Longhorn storage.
s
salmon-city-57654
07/05/2023, 5:33 PMHi @broad-knife-10346!
Thanks for your sharing. The various read pattern would affect the performance result for real cases, so we could do some simple tests for it. (like run some VM with containers and do some benchmarks)
Also, transferring your data from the existing SAN to Longhorn Storage should be OK. Maybe you could try with v1.1.2 or v1.2.0. v1.2.0 will release in recent.
Hope we can hear more feedback from you after that!
21 Views