utoronto, terraform: refine comments slightly

2i2c-org · Jan 10, 2024 · d6d89f5 · d6d89f5
1 parent d02bc52
commit d6d89f5
Showing 1 changed file with 17 additions and 6 deletions.
diff --git a/terraform/azure/projects/utoronto.tfvars b/terraform/azure/projects/utoronto.tfvars
@@ -1,3 +1,11 @@
+# IMPORTANT: Due to a restrictive network rule from storage.tf, we can't perform
+#            "terraform plan" or "terraform apply" without a workaround.
+#
+#            One known workaround is to allow your public IP temporarily as
+#            discussed in https://github.com/2i2c-org/infrastructure/issues/890#issuecomment-1879072422.
+#            This workaround is problematic as that may temporarily allow access
+#            to storage by other actors with the same IP.
+#
 tenant_id                      = "78aac226-2f03-4b4d-9037-b46d56c55210"
 subscription_id                = "ead3521a-d994-4a44-a68d-b16e35642d5b"
 resourcegroup_name             = "2i2c-utoronto-cluster"
@@ -16,17 +24,20 @@ node_pools = {
     {
       name : "core",
 
-      # FIXME: transition to "Standard_E2s_v5" nodes as they are large enough and
-      #        can more cheaply handle being forced to have 2-3 replicas for silly
-      #        reasons like three calico-typha pods. See
-      #        https://github.com/2i2c-org/infrastructure/issues/3592#issuecomment-1883269632.
+      # FIXME: Transition to "Standard_E2s_v5" nodes as they are large enough to
+      #        for the biggest workload (prometheus-server) and can handle high
+      #        availability requirements better.
+      #
+      #        We are currently forced to handle three calico-typha pods that
+      #        can't schedule on the same node, see https://github.com/2i2c-org/infrastructure/issues/3592#issuecomment-1883269632.
       #
       vm_size : "Standard_E4s_v3",
 
+      # core nodes doesn't need much disk space
       os_disk_size_gb : 40,
 
-      # FIXME: stop using persistent disks for the nodes, use the variable default
-      #        "Temporary" instead
+      # FIXME: Stop using persistent disks for the nodes, use the variable default
+      #        "Temporary" instead by removing this line.
       kubelet_disk_type : "OS",
 
       min : 1,