review comments

cluster-autoscaler/cloudprovider/azure/README.md

@@ -197,7 +197,8 @@ The `AZURE_ENABLE_VMSS_FLEX` environment variable enables VMSS Flex support. By
 |---------------------------|---------|-----------------------------------------|---------------------------|
 | enableVmssFlex            | false   | AZURE_ENABLE_VMSS_FLEX                  | enableVmssFlex            |
 
-The `AZURE_GET_VMSS_SIZE_REFRESH_PERIOD` environment variable defines in seconds how frequently to call GET VMSS API to fetch VMSS info per nodegroup instance. By default, value is 30 seconds.
+The AZURE_GET_VMSS_SIZE_REFRESH_PERIOD environment variable defines, in seconds, the frequency to call the GET VMSS API to fetch VMSS info per spot nodegroup instance. 
+By default, the value is 30 seconds. Currently, this API is only utilized for spot instances.
 
 | Config Name               | Default | Environment Variable                | Cloud Config File         |
 |---------------------------|---------|-------------------------------------|---------------------------|

cluster-autoscaler/cloudprovider/azure/azure_scale_set.go

@@ -39,7 +39,6 @@ var (
 	defaultVmssInstancesRefreshPeriod = 5 * time.Minute
 	vmssContextTimeout                = 3 * time.Minute
 	vmssSizeMutex                     sync.Mutex
-	defaultGetVmssSizeRefreshPeriod   = VmssSizeRefreshPeriodDefault * time.Second
 )
 
 const (
@@ -60,19 +59,29 @@ type ScaleSet struct {
 	maxSize int
 
 	enableForceDelete bool
-
-	sizeMutex sync.Mutex
-	curSize   int64
-
 	enableDynamicInstanceList bool
 
+
+	// curSize tracks (and caches) the number of VMs in this ScaleSet.
+	// It is periodically updated from vmss.Sku.Capacity, with VMSS itself coming
+	// either from azure.Cache (which periodically does VMSS.List)
+	// or from direct VMSS.Get (used for Spot).
+	curSize   int64
+	// lastSizeRefresh is the time curSize was last refreshed from vmss.Sku.Capacity.
+	// Together with sizeRefreshPeriod, it is used to determine if it is time to refresh curSize.
 	lastSizeRefresh          time.Time
+	// sizeRefreshPeriod is how often curSize is refreshed from vmss.Sku.Capacity.
+	// (Set from azureCache.refreshInterval = VmssCacheTTL or [defaultMetadataCache]refreshInterval = 1min)
 	sizeRefreshPeriod        time.Duration
+	// getVmssSizeRefreshPeriod is how often curSize should be refreshed in case VMSS.Get call is used (only spot instances).
+	// (Set from GetVmssSizeRefreshPeriod, if specified = get-vmss-size-refresh-period = 30s,
+	//   or override from autoscalerProfile.GetVmssSizeRefreshPeriod)
 	getVmssSizeRefreshPeriod time.Duration
 
 	instancesRefreshPeriod time.Duration
 	instancesRefreshJitter int
 
+	sizeMutex sync.Mutex
 	instanceMutex       sync.Mutex
 	instanceCache       []cloudprovider.Instance
 	lastInstanceRefresh time.Time