[Storage] Fix sporadic NFS server start failure by using dynamic port assignment for all NFS services: statd, statd_out, mountd, lockd, rquotad.

Author: gmarciani

CHANGELOG.md

@@ -3,6 +3,12 @@ aws-parallelcluster-cookbook CHANGELOG
 
 This file is used to list changes made in each version of the AWS ParallelCluster cookbook.
 
+3.16.0
+------
+
+**CHANGES**
+- Reduce the risk of NFS server start failure by using dynamic port assignment for all NFS services: statd, statd_out, mountd, lockd, rquotad.
+
 3.15.0
 ------
 

cookbooks/aws-parallelcluster-environment/attributes/environment.rb

@@ -2,6 +2,16 @@
 
 default['cluster']['proxy'] = 'NONE'
 
+# Use dynamic port assignment for all NFS services to avoid EADDRINUSE collisions
+# with the Linux ephemeral port range (32768-60999). The upstream NFS cookbook defaults
+# (32765-32769) overlap with or sit adjacent to this range, causing sporadic boot failures.
+# PCluster doesn't need static NFS ports — nodes communicate within a VPC using security groups.
+default['nfs']['port']['statd'] = 0
+default['nfs']['port']['statd_out'] = 0
+default['nfs']['port']['mountd'] = 0
+default['nfs']['port']['lockd'] = 0
+default['nfs']['port']['rquotad'] = 0
+
 # For performance, set NFS threads to min(256, max(8, num_cores * 4))
 default['cluster']['nfs']['threads'] = [[node['cpu']['cores'].to_i * 4, 8].max, 256].min
 

cookbooks/aws-parallelcluster-environment/spec/unit/resources/nfs_spec.rb

@@ -18,6 +18,21 @@ def self.configure(chef_run)
   end
 end
 
+describe 'nfs:port_overrides' do
+  # Verify that all NFS service ports are overridden to 0 (dynamic assignment)
+  # to avoid EADDRINUSE collisions with the Linux ephemeral port range (32768-60999).
+  # The upstream NFS cookbook defaults (32765-32769) overlap with or sit adjacent to
+  # this range, causing sporadic NFS server start failures during boot.
+  cached(:chef_run) { ChefSpec::SoloRunner.new.converge('aws-parallelcluster-environment') }
+  cached(:node) { chef_run.node }
+
+  %w(statd statd_out mountd lockd rquotad).each do |service|
+    it "overrides nfs port for #{service} to 0 (dynamic)" do
+      expect(node['nfs']['port'][service]).to eq(0)
+    end
+  end
+end
+
 describe 'nfs:setup' do
   for_all_oses do |platform, version|
     context "on #{platform}#{version}" do