Avoid timeout in capture-system-logs due to df command

The `df` command can stall indefinitely on stale NFS mounts,
causing the playbook to time out. This leads to the entire job
failing with POST_FAILURE status, discarding controller logs
and impacting troubleshooting.

This patch changes `capture-system-logs` to run `df` with a 60s
timeout to prevent hangs from stale NFS mounts. If 'df' times out,
the mount output may help debug which NFS share is unresponsive.

Change-Id: Ife3945802c93bd77d60b60e433ea09aade38a522
Signed-off-by: Fernando Ferraz <fernandoperches@gmail.com>

.zuul.yaml

@@ -446,6 +446,7 @@
         /etc/sudoers.d: logs
         '{{ stage_dir }}/iptables.txt': logs
         '{{ stage_dir }}/df.txt': logs
+        '{{ stage_dir }}/mount.txt': logs
         '{{ stage_dir }}/pip2-freeze.txt': logs
         '{{ stage_dir }}/pip3-freeze.txt': logs
         '{{ stage_dir }}/dpkg-l.txt': logs

roles/capture-system-logs/tasks/main.yaml

@@ -4,7 +4,13 @@
     executable: /bin/bash
     cmd: |
       sudo iptables-save > {{ stage_dir }}/iptables.txt
-      df -h > {{ stage_dir }}/df.txt
+
+      # NOTE(sfernand): Run 'df' with a 60s timeout to prevent hangs from
+      # stale NFS mounts.
+      timeout -s 9 60s df -h > {{ stage_dir }}/df.txt || true
+      # If 'df' times out, the mount output helps debug which NFS share
+      # is unresponsive.
+      mount > {{ stage_dir }}/mount.txt
 
       for py_ver in 2 3; do
           if [[ `which python${py_ver}` ]]; then