Fix possible deadlock for provisioning bare-metal clusters

* Closes #235
poseidon · Jun 15, 2018 · ed0b781 · ed0b781
1 parent 51906bf
commit ed0b781
Show file tree

Hide file tree

Showing 3 changed files with 37 additions and 1 deletion.
diff --git a/CHANGES.md b/CHANGES.md
@@ -4,7 +4,11 @@ Notable changes between versions.
 
 ## Latest
 
-* Update etcd from v3.3.6 to v3.3.7
+* Update etcd from v3.3.6 to v3.3.7 ([#243](https://github.com/poseidon/typhoon/pull/243))
+
+#### Bare-Metal
+
+* Fix possible deadlock when provisioning clusters larger than 10 nodes ([#244](https://github.com/poseidon/typhoon/pull/244)) 
 
 #### Addons
 

diff --git a/bare-metal/container-linux/kubernetes/ssh.tf b/bare-metal/container-linux/kubernetes/ssh.tf
@@ -2,6 +2,14 @@
 resource "null_resource" "copy-controller-secrets" {
   count = "${length(var.controller_names)}"
 
+  # Without depends_on, remote-exec could start and wait for machines before
+  # matchbox groups are written, causing a deadlock.
+  depends_on = [
+    "matchbox_group.install",
+    "matchbox_group.controller",
+    "matchbox_group.worker",
+  ]
+
   connection {
     type    = "ssh"
     host    = "${element(var.controller_domains, count.index)}"
@@ -70,6 +78,14 @@ resource "null_resource" "copy-controller-secrets" {
 resource "null_resource" "copy-worker-secrets" {
   count = "${length(var.worker_names)}"
 
+  # Without depends_on, remote-exec could start and wait for machines before
+  # matchbox groups are written, causing a deadlock.
+  depends_on = [
+    "matchbox_group.install",
+    "matchbox_group.controller",
+    "matchbox_group.worker",
+  ]
+
   connection {
     type    = "ssh"
     host    = "${element(var.worker_domains, count.index)}"

diff --git a/bare-metal/fedora-atomic/kubernetes/ssh.tf b/bare-metal/fedora-atomic/kubernetes/ssh.tf
@@ -2,6 +2,14 @@
 resource "null_resource" "copy-controller-secrets" {
   count = "${length(var.controller_names)}"
 
+  # Without depends_on, remote-exec could start and wait for machines before
+  # matchbox groups are written, causing a deadlock.
+  depends_on = [
+    "matchbox_group.install",
+    "matchbox_group.controller",
+    "matchbox_group.worker",
+  ]
+
   connection {
     type    = "ssh"
     host    = "${element(var.controller_domains, count.index)}"
@@ -68,6 +76,14 @@ resource "null_resource" "copy-controller-secrets" {
 resource "null_resource" "copy-worker-secrets" {
   count = "${length(var.worker_names)}"
 
+  # Without depends_on, remote-exec could start and wait for machines before
+  # matchbox groups are written, causing a deadlock.
+  depends_on = [
+    "matchbox_group.install",
+    "matchbox_group.controller",
+    "matchbox_group.worker",
+  ]
+
   connection {
     type    = "ssh"
     host    = "${element(var.worker_domains, count.index)}"