Service Fabric Node Balancing doesn't trigger -
i attempt test service fabric node balancing mechanism on own private cluster , can’t figure out why doesn’t work. below steps took:
configure cluster resource manager timer periodically check balancing state 5s, set balancing threshold metric “memoryusage” defined 2 , define node capacity metric 1200. clustermanifest looks like: clustermanifest
create 2 stateless application: 1 called “nodebalancingtestappwith500” defaultload 200, weight high , instance count 5 , other 1 called “nodebalanceteststatelessservice” static load 400, weight high , instance count 1 publish both cluster. applicationmanifest looks like: applicationmanifest
after publish 2 apps cluster, load in 5 nodes 200,200, 600, 200 , 200. expect see app “nodebalancingtestappwith500” defaultload 200 in node 3 should moved node after resource manager kicks in. after balancing , should 200,400,400,200,200 this?
could tell why node balancing doesn’t work?
seems there no capacity violation considering you've defined 1200 node capacity given metric. if expand nodes in sf explorer - show you? take @ "is capacity violation" , "node remaining capacity" columns.
other that, set "metricbalancingthresholds" sure current ratio (600/200) suggest sf cluster imbalanced -
a balancing threshold main control triggering rebalancing. balancing threshold metric ratio. if load metric on loaded node divided amount of load on least loaded node exceeds metric's balancingthreshold, cluster imbalanced. result balancing triggered next time cluster resource manager checks.
Comments
Post a Comment