azure - Turning off ServiceFabric clusters overnight -


we working on application processes excel files , spits off output. availability not big requirement.

can turn vm sets off during night , turn them on again in morning? kind of setup work service fabric? if so, there way schedule it?

thank replying. i've got chance talk microsoft azure rep , documented conversation in here community sake.

response initial question

a service fabric cluster must maintain minimum number of primary node types in order system services maintain quorum , ensure health of cluster. can see more reliability level , instance count @ https://azure.microsoft.com/en-gb/documentation/articles/service-fabric-cluster-capacity/. such, stopping of vms cause service fabric cluster go quorum loss. possible bring nodes , service fabric automatically recover quorum loss, not guaranteed , cluster may never able recover.

however, if not need save state in cluster may easier delete , recreate entire cluster (the entire azure resource group) every day. creating new cluster scratch deploying new resource group takes less half hour, , can automated using powershell deploy arm template. https://azure.microsoft.com/en-us/documentation/articles/service-fabric-cluster-creation-via-arm/ shows how setup arm template , deploy using powershell. can additionally use fixed domain name or static ip address clients don’t have reconfigured connect cluster. if have need maintain other resources such storage account configure arm template delete vm scale set , sf cluster resource while keeping network, load balancer, storage accounts, etc.

q)is there better way stop/start vms rather directly scale set?

if want stop vms in order save cost, starting/stopping vms directly scale set option.

q) can primary set cheapest vms can find , add secondary set powerful vms can turn on , off?

yes, possible create 2 node types – primary small/cheap, , ‘worker’ larger size – , set placement constraints on application deploy larger size vms. however, if service fabric service storing state still run similar problem once lose quorum (below 3 replicas/nodes) of worker vm there no guarantee sf service come of state maintained. in case cluster still fine since primary nodes running, service’s state may in unknown replication state.

i think have few options:

  1. instead of storing state within service fabric’s reliable collections, instead store state externally azure storage or sql azure. can optionally use redis cache or service fabric’s reliable collections in order maintain faster read-cache, make sure writes persisted external store. way can freely delete , recreate cluster @ time want.
  2. use service fabric backup/restore in order maintain state, , delete entire resource group or cluster overnight , recreate , restore state in morning. backup/restore duration depend entirely on how data storing , export backup.
  3. utilize such azure batch. service fabric not designed temporary high capacity compute platform can started , stopped regularly, if goal may want @ hpc platform such azure batch offers native capabilities burst compute capacity.

Comments

Popular posts from this blog

unity3d - Rotate an object to face an opposite direction -

angular - Is it possible to get native element for formControl? -

javascript - Why jQuery Select box change event is now working? -