COA Troubleshooting & Problem Resolution 3 — Questions and Answers
Question 1: A Cinder backup operation completes but restoring from it fails with 'Invalid backup'. Which component's logs should be checked first?
- /var/log/cinder/api.log
- /var/log/cinder/backup.log (Correct answer)
- /var/log/cinder/scheduler.log
- /var/log/cinder/volume.log
Correct answer: /var/log/cinder/backup.log
The cinder-backup service handles backup and restore operations, so its log is the authoritative source for backup integrity failures.
Question 2: After a power failure, several instances show as 'SHUTOFF' in Nova but are actually running on the hypervisor. Which command reconciles Nova's view with the actual hypervisor state?
- nova reset-state --active <instance-id> (Correct answer)
- openstack server reboot <instance-id>
- nova-manage vm reset_state
- openstack compute service set --enable
Correct answer: nova reset-state --active <instance-id>
'nova reset-state --active' forces Nova's database record to match the hypervisor reality without touching the running VM.
Question 3: An OVS (Open vSwitch) agent on a compute node shows as 'DOWN' in Neutron. What is the correct sequence of steps to recover it?
- Reboot the compute node
- Restart the neutron-openvswitch-agent service, then verify with openstack network agent list (Correct answer)
- Delete and recreate the agent entry in Neutron DB
- Run neutron-db-manage upgrade head
Correct answer: Restart the neutron-openvswitch-agent service, then verify with openstack network agent list
Restarting the OVS agent service re-registers it with the Neutron server; confirming with 'network agent list' verifies successful recovery.
Question 4: A user cannot create snapshots of a running instance, receiving 'Snapshot not supported'. What is the most likely cause?
- The Glance backend storage is full
- The compute node's hypervisor driver does not support live snapshots
- Cinder quota has been exceeded
- The image is based on a volume-backed instance (Correct answer)
Correct answer: The image is based on a volume-backed instance
Volume-backed (boot-from-volume) instances do not support Nova snapshots the same way; users must create a Cinder volume snapshot instead.
Question 5: Which log file on the controller node should be inspected when Glance image uploads fail with HTTP 500 errors?
- /var/log/glance/registry.log
- /var/log/glance/api.log (Correct answer)
- /var/log/keystone/keystone.log
- /var/log/swift/proxy-server.log
Correct answer: /var/log/glance/api.log
The glance-api.log records all API request errors including upload failures, making it the first file to inspect for HTTP 500 responses.
Question 6: A tenant complains that DNS resolution is broken inside their instances. The DHCP agent is running and leases are being issued. What should you check next?
- The neutron-metadata-agent status
- The dns_nameservers attribute of the subnet (Correct answer)
- The router's static routes
- The Nova console log for DNS errors
Correct answer: The dns_nameservers attribute of the subnet
When DHCP works but DNS fails, the subnet's dns_nameservers configuration is the most common culprit — missing or incorrect entries cause broken resolution.
Question 7: You need to diagnose why a specific tenant's network traffic is being dropped at the compute node level. Which tool directly inspects OVS flow rules on the host?
- neutron-debug probe-create
- ovs-ofctl dump-flows br-int (Correct answer)
- ip netns exec <ns> tcpdump
- openstack port show <port-id>
Correct answer: ovs-ofctl dump-flows br-int
'ovs-ofctl dump-flows br-int' dumps all OpenFlow rules on the integration bridge, letting you trace where tenant traffic is being dropped.
COA Troubleshooting & Problem Resolution 3
A Cinder backup operation completes but restoring from it fails with 'Invalid backup'.
Which component's logs should be checked first?