Capture QEMU core dumps when possible

Some of the hardest-to-debug issues are qemu crashes deep in a nova
workflow that can't be reproduced locally. This adds a post task to
the playbook so that we capture the most recent qemu core dump, if
there is one.

Change-Id: I48a2ea883325ca920b7e7909edad53a9832fb319

Author: kk7ds

.zuul.yaml

@@ -377,6 +377,7 @@
         '{{ devstack_log_dir }}/devstacklog.txt.summary': logs
         '{{ devstack_log_dir }}/tcpdump.pcap': logs
         '{{ devstack_log_dir }}/worlddump-latest.txt': logs
+        '{{ devstack_log_dir }}/qemu.coredump': logs
         '{{ devstack_full_log}}': logs
         '{{ stage_dir }}/verify_tempest_conf.log': logs
         '{{ stage_dir }}/performance.json': logs

playbooks/post.yaml

@@ -17,6 +17,12 @@
         dest: "{{ stage_dir }}/verify_tempest_conf.log"
         state: hard
       when: tempest_log.stat.exists
+    - name: Capture most recent qemu crash dump, if any
+      shell:
+        executable: /bin/bash
+        cmd: |
+          coredumpctl -o {{ devstack_log_dir }}/qemu.coredump dump /usr/bin/qemu-system-x86_64
+      ignore_errors: yes
   roles:
     - export-devstack-journal
     - apache-logs-conf