Merge branch 'main' into refactor/firewalld

stackhpc · Mar 7, 2024 · 6231b20 · 6231b20
2 parents bab676f + 6ca97fe
commit 6231b20
Show file tree

Hide file tree

Showing 10 changed files with 20 additions and 4 deletions.
diff --git a/ansible/roles/filebeat/tasks/install.yml b/ansible/roles/filebeat/tasks/install.yml
@@ -15,3 +15,4 @@
 - name: Reload filebeat unit file
   command: systemctl daemon-reload
   when: _filebeat_unit.changed
+  become: true
diff --git a/ansible/roles/hpctests/README.md b/ansible/roles/hpctests/README.md
@@ -33,6 +33,7 @@ Role Variables
 - `hpctests_hpl_arch`: Optional, default 'linux64'. Arbitrary architecture name for HPL build. HPL is compiled on the first compute node of those selected (see `hpctests_nodes`), so this can be used to create different builds for different types of compute node.
 
 The following variables should not generally be changed:
+- `hpctests_pre_cmd`: Optional. Command(s) to include in sbatch templates before module load commands.
 - `hpctests_pingmatrix_modules`: Optional. List of modules to load for pingmatrix test. Defaults are suitable for OpenHPC 2.x cluster using the required packages.
 - `hpctests_pingpong_modules`: As above but for pingpong test.
 - `hpctests_pingpong_plot`: Whether to plot pingpong results. Default `yes`.

diff --git a/ansible/roles/hpctests/defaults/main.yml b/ansible/roles/hpctests/defaults/main.yml
@@ -1,5 +1,6 @@
 ---
 hpctests_rootdir:
+hpctests_pre_cmd: ''
 hpctests_pingmatrix_modules: [gnu12 openmpi4]
 hpctests_pingpong_modules: [gnu12 openmpi4 imb]
 hpctests_pingpong_plot: yes

diff --git a/ansible/roles/hpctests/templates/hpl-build.sh.j2 b/ansible/roles/hpctests/templates/hpl-build.sh.j2
@@ -8,7 +8,7 @@
 {%if hpctests_nodes is defined %}#SBATCH --nodelist={{ hpctests_computes.stdout_lines[0] }}{% endif %}
 
 echo HPL arch: {{ hpctests_hpl_arch }}
-
+{{ hpctests_pre_cmd }}
 module load {{ hpctests_hpl_modules | join(' ' ) }}
 make arch={{ hpctests_hpl_arch }} clean_arch_all
 make arch={{ hpctests_hpl_arch }}
diff --git a/ansible/roles/hpctests/templates/hpl-solo.sh.j2 b/ansible/roles/hpctests/templates/hpl-solo.sh.j2
@@ -15,6 +15,6 @@ echo SLURM_JOB_NODELIST: $SLURM_JOB_NODELIST
 echo SLURM_JOB_ID: $SLURM_JOB_ID
 echo UCX_NET_DEVICES: $UCX_NET_DEVICES
 echo HPL arch: {{ hpctests_hpl_arch }}
-
+{{ hpctests_pre_cmd }}
 module load {{ hpctests_hpl_modules | join(' ' ) }}
 mpirun ./xhpl-{{ hpctests_hpl_arch }}
diff --git a/ansible/roles/hpctests/templates/pingmatrix.sh.j2 b/ansible/roles/hpctests/templates/pingmatrix.sh.j2
@@ -12,6 +12,7 @@ export UCX_NET_DEVICES={{ hpctests_ucx_net_devices }}
 echo SLURM_JOB_NODELIST: $SLURM_JOB_NODELIST
 echo SLURM_JOB_ID: $SLURM_JOB_ID
 echo UCX_NET_DEVICES: $UCX_NET_DEVICES
+{{ hpctests_pre_cmd }}
 module load {{ hpctests_pingmatrix_modules | join(' ' ) }}
 
 mpicc -o nxnlatbw mpi_nxnlatbw.c

diff --git a/ansible/roles/hpctests/templates/pingpong.sh.j2 b/ansible/roles/hpctests/templates/pingpong.sh.j2
@@ -12,6 +12,7 @@ export UCX_NET_DEVICES={{ hpctests_ucx_net_devices }}
 echo SLURM_JOB_NODELIST: $SLURM_JOB_NODELIST
 echo SLURM_JOB_ID: $SLURM_JOB_ID
 echo UCX_NET_DEVICES: $UCX_NET_DEVICES
+{{ hpctests_pre_cmd }}
 module load {{ hpctests_pingpong_modules | join(' ' ) }}
 
 #srun --mpi=pmi2 IMB-MPI1 pingpong # doesn't work in ohpc v2.1

diff --git a/environments/.stackhpc/LEAFCLOUD.pkrvars.hcl b/environments/.stackhpc/LEAFCLOUD.pkrvars.hcl
@@ -0,0 +1,11 @@
+flavor = "en1.xsmall"
+use_blockstorage_volume = true
+volume_size = 12 # GB. Compatible with SMS-lab's general.v1.tiny 
+image_disk_format = "qcow2"
+networks = ["909e49e8-6911-473a-bf88-0495ca63853c"] # slurmapp-ci
+source_image_name = "openhpc-230804-1754-80b8d714" # https://github.com/stackhpc/ansible-slurm-appliance/pull/298
+fatimage_source_image_name = "Rocky-8-GenericCloud-Base-8.9-20231119.0.x86_64.qcow2"
+ssh_keypair_name = "slurm-app-ci"
+ssh_private_key_file = "~/.ssh/id_ed25519"
+security_groups = ["default", "SSH"]
+floating_ip_network = "external"
diff --git a/environments/.stackhpc/inventory/group_vars/all/bastion.yml b/environments/.stackhpc/inventory/group_vars/all/bastion.yml
@@ -4,7 +4,7 @@ bastion_config:
     user: slurm-app-ci
     ip: 128.232.222.183
   LEAFCLOUD:
-    user: rocky
+    user: slurm-app-ci
     ip: 195.114.30.222
 # NB: The bastion_{user,ip} variables are used directly in the CI workflow too
 bastion_user: "{{ bastion_config[ci_cloud].user }}"

diff --git a/environments/.stackhpc/terraform/LEAFCLOUD.tfvars b/environments/.stackhpc/terraform/LEAFCLOUD.tfvars
@@ -1,6 +1,6 @@
 cluster_net = "slurmapp-ci"
 cluster_subnet = "slurmapp-ci"
 control_node_flavor = "ec1.medium" # small ran out of memory, medium gets down to ~100Mi mem free on deployment
-other_node_flavor = "ec1.small"
+other_node_flavor = "en1.xsmall"
 state_volume_type = "unencrypted"
 home_volume_type = "unencrypted"