This message was deleted.
# longhorn-storagea
adamant-kite-43734
04/08/2024, 10:30 AMThis message was deleted.
n
nice-tent-65195
04/08/2024, 11:10 AMHello!
After reboot Longhorn normally recalculates the checksums and copies the differences to the restarted replica server. So I think that's perfectly normal
c
cool-architect-86201
04/08/2024, 12:42 PMThank you @nice-tent-65195, Also, could you please explain the scheduling strategy of Longhore? After successfully scaling up, I installed a large number of blockchain nodes (applied for many volumes), but it seems that most of them went to worker04 node
n
nice-tent-65195
04/08/2024, 1:43 PMPerhaps the rke-worker-04 node was the first node which went to Ready after the reboot. If that's true, then all the 'Degraded' volumes started a rebuild process on the the 04 node, because the other nodes were unavailable. But that's just my guess
nice-tent-65195
04/08/2024, 1:43 PMYou can read all the details about scheduling here: https://longhorn.io/docs/1.6.0/nodes-and-volumes/nodes/scheduling/#scheduling-policy
nice-tent-65195
04/08/2024, 1:46 PMOh, forget it, I see you're only using one replica/volume.
nice-tent-65195
04/08/2024, 1:48 PMThis can be dangerous because if you lose just one node, you can lose important data with it.
🎯 1
c
cool-architect-86201
04/09/2024, 2:53 AMThank you for your reminder @nice-tent-65195. We intentionally set it to 1 because we are all blockchain nodes, and the data can be downloaded at any time. We deploy multiple pods, each applying for PVC, in order to complete rolling upgrades.
👍 1
l
late-needle-80860
04/09/2024, 8:32 PM@cool-architect-86201 when you’re saying blockchain node … what do you mean more exactly? Some crypto related workloads?
c
cool-architect-86201
04/10/2024, 3:41 AMHi @late-needle-80860 Yes, we have deployed a lot of blockchain nodes(pods) API for customer queries. For example, Bitcoin, Ethereum, Polkadot, Cosmos, etc.
cool-architect-86201
04/10/2024, 3:44 AMHi @nice-tent-65195 I'd like to ask another question regarding which network card Longhore traffic is using. I just raised a related issue on RKE2（https://github.com/rancher/rke2/discussions/5699）. Each of our devices has two network cards, both running at 10Gbps. On Ubuntu, they are labeled as eno1 and eno2. Among them, eno1 is our public network address, and eno2 is our internal network address.
Now, there is an issue. When my worker nodes join the master, they write the master's eno2 address. But now I suspect they are communicating with each other using the eno1 network card. This is causing higher traffic overhead on eno1, resulting in unnecessary expenses.
I found through "describe node" that all addresses are eno1's IPv4 and IPv6 addresses, identified as "Internal."
My question is, does Longhore communication use the addresses of K8s Nodes? That is, the current public network addresses.
l
late-needle-80860
04/10/2024, 6:07 PMI don’t know about RKE … But, Longhorn supports using multiple nics .. one for backbone communication and so on. Look in the docs
2 Views