Reputation: 27
We have a problem statement i.e. We are using Azure Service Fabric for our production. We have service fabric with Silver Tier. Our issue is when a single instance gets Spike i.e. due to High CPU utilization and Memory. Load balancer is unable to transfer request to other nodes. Single node get 90 percent utilization and we are even unable to RDP that node during that time. I have seen articles from Microsoft about adding placement constraints. Still that didn't work either. We are unable to apply rules to loadbalancer as we have integrated APIM with Service Fabric. I had multiple calls with Microsoft Still didn't get appropriate solution which could work. I need a solution to my problem.
I know we have issue in one of our services we are already working on it but we need SF to handle this scenario as well.
Upvotes: 0
Views: 653
Reputation: 311
If one or more of your services generates CPU / memory spikes (and not a consistent high utilization) then it will be very hard to balance such behavior.
Anyway, you can do two things to mitigate it:
Use resource governance to restrict the amount of CPU and memory that this problematic service can consume
Microsoft released FabricObserver which can be used to extend the monitoring of our SF cluster. You can have a look and see how you can leverage AppObserver to report CPU and memory usages of a single service (process) as LoadMetrics and use it to balance the cluster
Upvotes: 0