aws-samples/aws-eda-slurm-cluster

HeadNode fails to configure due to ansible change. on_head_node_configured.sh fails as ansible has deprecated ansible.builtin.include

gwolski opened this issue · 2 comments

gwolski commented

Using latest version of aws-eda-slurm-cluster as of May 15.

Headnode fails to configure. cloud init fails. Reproduce by deploying new cluster with latest ansible

Tracked things down so far to the shell script on_head_node_configured.sh. Error message is:

<13>Jun 10 16:48:18 on_head_node_configured.sh: TASK [all : Create /var/lib/cloud/scripts/per-boot/90_mount_ssds.bash] *********
<13>Jun 10 16:48:18 on_head_node_configured.sh: ok: [local]
<13>Jun 10 16:48:18 on_head_node_configured.sh:
<13>Jun 10 16:48:18 on_head_node_configured.sh: TASK [all : Execute /var/lib/cloud/scripts/per-boot/90_mount_ssds.bash] ********
<13>Jun 10 16:48:18 on_head_node_configured.sh: ok: [local]
<13>Jun 10 16:48:18 on_head_node_configured.sh:
<13>Jun 10 16:48:18 on_head_node_configured.sh: TASK [all : Give /tmp write permissions] ***************************************
<13>Jun 10 16:48:18 on_head_node_configured.sh: ok: [local]
<13>Jun 10 16:48:18 on_head_node_configured.sh: ERROR! [DEPRECATED]: ansible.builtin.include has been removed. Use include_tasks or import_tasks instead. This feature was removed from ansible-core in a release after 2023-05-16. Please update your playbooks.
<13>Jun 10 16:48:18 on_head_node_configured.sh: + on_exit
<13>Jun 10 16:48:18 on_head_node_configured.sh: + rc=1
<13>Jun 10 16:48:18 on_head_node_configured.sh: + set +e
<13>Jun 10 16:48:18 on_head_node_configured.sh: + [[ 1 -ne 0 ]]
<13>Jun 10 16:48:18 on_head_node_configured.sh: + [[ : != : ]]

include is used in
/opt/slurm/config/ansible/playbooks/roles/ParallelClusterHeadNode/tasks/main.yml

gwolski commented

Changing tasks/main.yml to call ansible.builtin.import_tasks allows the on_head_node_configured.sh to run.
Here is the updated tasks/main.yml. I do not know enough about the replacement tasks to be 100% sure this is the best/right:

  • { ansible.builtin.import_tasks: config-users-groups.yml, tags: users-groups }
  • { ansible.builtin.import_tasks: config-high-throughput.yml, tags: high-throughput }
  • { ansible.builtin.import_tasks: config-submitter-access.yml, tags: submitter }
  • { ansible.builtin.import_tasks: config-sshd.yml, tags: sshd }
  • { ansible.builtin.import_tasks: config-slurmrestd.yml, tags: slurmrestd }
  • { ansible.builtin.import_tasks: config-licenses.yml, tags: licenses }
  • { ansible.builtin.import_tasks: config-slurmdb-accounts.yml, tags: accounts }

Will attempt to build a new cluster.

gwolski commented

New cluster is working, build completed cleanly. Please do not trust my selection of ansible import_tasks as replacement for original "include" command.