[SlurmTopo] Solve the update failures by defining when to update topology.conf (#3009)

Co-authored-by: Himani Anil Deshpande <[email protected]>

Author: himani2411

cookbooks/aws-parallelcluster-slurm/attributes/slurm_attributes.rb

@@ -28,3 +28,4 @@
 
 # Block Topology Plugin
 default['cluster']['slurm']['block_topology']['force_configuration'] = false
+default['cluster']['p6egb200_block_sizes'] = nil

cookbooks/aws-parallelcluster-slurm/resources/block_topology/partial/_block_topology_common.rb

@@ -63,10 +63,8 @@ def topology_generator_command_args
   if node['cluster']['p6egb200_block_sizes'].nil? && are_queues_updated? && ::File.exist?("#{node['cluster']['slurm']['install_dir']}/etc/topology.conf")
     # If topology.conf exist and Capacity Block is removed, we cleanup
     " --cleanup"
-  elsif node['cluster']['p6egb200_block_sizes'].nil? && !are_queues_updated?
-    # We do nothing if p6e-gb200 is not used and queues are not updated
-    nil
-  else
+  elsif !node['cluster']['p6egb200_block_sizes'].nil?
+    # We add/update topology.conf if p6egb200_block_sizes is not null
     " --block-sizes #{node['cluster']['p6egb200_block_sizes']}"
   end
 end

cookbooks/aws-parallelcluster-slurm/spec/unit/resources/block_topology_spec.rb

@@ -154,13 +154,26 @@ def self.update(chef_run)
       cached(:chef_run) do
         runner(platform: platform, version: version, step_into: ['block_topology']) do |node|
           node.override['cluster']['p6egb200_block_sizes'] = nil
+          node.override['cluster']['slurm']['install_dir'] = slurm_install_dir
         end
       end
       cached(:resource) do
         ConvergeBlockTopology.update(chef_run)
         chef_run.find_resource('block_topology', 'update')
       end
 
+      context "when queues are updated and topolog.conf does exists" do
+        before do
+          allow_any_instance_of(Object).to receive(:are_queues_updated?).and_return(true)
+          allow(File).to receive(:exist?).with("#{slurm_install_dir}/etc/topology.conf").and_return(true)
+          chef_run.node.override['cluster']['p6egb200_block_sizes'] = nil
+        end
+
+        it 'returns cleanup' do
+          expect(resource.topology_generator_command_args).to eq(" --cleanup")
+        end
+      end
+
       context "when queues are not updated and topolog.conf does not exists" do
         before do
           allow_any_instance_of(Object).to receive(:are_queues_updated?).and_return(false)
@@ -172,6 +185,18 @@ def self.update(chef_run)
         end
       end
 
+      context "when queues are updated and topolog.conf does not exists" do
+        before do
+          allow_any_instance_of(Object).to receive(:are_queues_updated?).and_return(true)
+          allow(File).to receive(:exist?).with("#{slurm_install_dir}/etc/topology.conf").and_return(false)
+          chef_run.node.override['cluster']['p6egb200_block_sizes'] = block_sizes
+        end
+
+        it 'returns block-sizes argument' do
+          expect(resource.topology_generator_command_args).to eq(" --block-sizes #{block_sizes}")
+        end
+      end
+
       context "when block sizes is not nil" do
         before do
           chef_run.node.override['cluster']['p6egb200_block_sizes'] = block_sizes