Service Fabric 群集状态“升级服务无法访问”

Question

I Had SF cluster made of 3 Standard A0 nodes. I scaled cluster in to 1 node and understood that this was bad idea because nothing was working in this state (even SF explorer was not working) Then I scaled it out back to 3 nodes and restarted Primary scaleser. Now all nodes in scaleset are up and running but SF cluster status is "Upgrade service unreachable". I saw similar question Service Fabric Status: Upgrade service unreachable where was recommended to scale nodes up to D2 but this hasn't solve my problem. I have connected to one node via RDP and are some Event logs:

EventLog -> Applications and Service Logs -> Microsoft Service Fabric -> Operational:

Node name: _SSService_0 has failed to open with upgrade domain: 0, fault domain: fd:/0, address: 10.0.0.4, hostname: SSService000000, isSeedNode: true, versionInstance: 5.6.210.9494:3, id: d9e8bae2d4d8116bfefb989b95e91f7b, dca instance: 131405546580494698, error: FABRIC_E_TIMEOUT

EventLog -> Applications and Service Logs -> Microsoft Service Fabric -> Admin:

client-10.0.0.4:19000/10.0.0.4:19000: error = 2147943625, failureCount=487. Filter by (type~Transport.St && ~"(?i)10.0.0.4:19000") to get listener lifecycle. Connect failure is expected if listener was never started, or listener/its process was stopped before/during connecting.

Answer 1

If you are scaling down the cluster by resizing VM scale set to 1 you're basically destroying the cluster because it requires a minimum of 3 nodes by design. Therefore the only way is to recreate it again from scratch.

If you need a tiny cluster consisting of just 1 node (like for testing purposes) there is a way in Azure now to create a single node cluster, but you won't be able to scale it as it's a special case not for production use.

Answer 2

Upgrade service unreachable this happens if the number of active VM or node of the cluster become 0 anyhow. In my case, his happened by restarting all the VM at a time. In this state, the nodes are available and running but they have been disconnected from the cluster.

I resolved this, by deallocating and restarting the node from Virtual machine Scale set .