From b434bdab149ebe24510df2d4a55dd9ba9746cd40 Mon Sep 17 00:00:00 2001
From: Steve Brasier <33413598+sjpb@users.noreply.github.com>
Date: Thu, 7 Mar 2024 10:31:00 +0000
Subject: [PATCH 1/3] fix filebeat container pull PR#351 (#373)

---
 ansible/roles/filebeat/tasks/install.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/ansible/roles/filebeat/tasks/install.yml b/ansible/roles/filebeat/tasks/install.yml
index 8e64722ec..6514e3028 100644
--- 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

From 17f2b2dcb6b2255db587c74e86bba475165c8195 Mon Sep 17 00:00:00 2001
From: Steve Brasier <33413598+sjpb@users.noreply.github.com>
Date: Thu, 7 Mar 2024 10:31:14 +0000
Subject: [PATCH 2/3] add hpctests_pre_cmd option (#375)

---
 ansible/roles/hpctests/README.md                  | 1 +
 ansible/roles/hpctests/defaults/main.yml          | 1 +
 ansible/roles/hpctests/templates/hpl-build.sh.j2  | 2 +-
 ansible/roles/hpctests/templates/hpl-solo.sh.j2   | 2 +-
 ansible/roles/hpctests/templates/pingmatrix.sh.j2 | 1 +
 ansible/roles/hpctests/templates/pingpong.sh.j2   | 1 +
 6 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/ansible/roles/hpctests/README.md b/ansible/roles/hpctests/README.md
index ee37791ec..5c5415158 100644
--- 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
index 280fd454e..30ddd8952 100644
--- 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
index f243a08f7..9300231e8 100644
--- 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
index cc7d2b4dd..6178cd60e 100644
--- 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
index 17fb3fd6a..d886e9ac8 100644
--- 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
index e74e52539..4dc2eebd5 100644
--- 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

From 6ca97fede4fa8bd13139bd7e3ddebb55f457e290 Mon Sep 17 00:00:00 2001
From: Steve Brasier <33413598+sjpb@users.noreply.github.com>
Date: Thu, 7 Mar 2024 10:35:53 +0000
Subject: [PATCH 3/3] Improve leafcloud CI support (#374)

* use proper leafcloud jumphost user

* use cheaper leafcloud flavor

* add leafcloud packer vars file
---
 environments/.stackhpc/LEAFCLOUD.pkrvars.hcl          | 11 +++++++++++
 .../.stackhpc/inventory/group_vars/all/bastion.yml    |  2 +-
 environments/.stackhpc/terraform/LEAFCLOUD.tfvars     |  2 +-
 3 files changed, 13 insertions(+), 2 deletions(-)
 create mode 100644 environments/.stackhpc/LEAFCLOUD.pkrvars.hcl

diff --git a/environments/.stackhpc/LEAFCLOUD.pkrvars.hcl b/environments/.stackhpc/LEAFCLOUD.pkrvars.hcl
new file mode 100644
index 000000000..2f325d754
--- /dev/null
+++ 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
index 626f9ff5c..94287827c 100644
--- 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
index ec152ca42..5e73896c8 100644
--- 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"