Skip to content

Latest commit

 

History

History
50 lines (38 loc) · 2.42 KB

README.md

File metadata and controls

50 lines (38 loc) · 2.42 KB
Raw

Troubleshooting

Note

By default, logs are published to the node in which the EFS mount occurs. This node can be found by examining the output of kubectl describe pod efs-app, where efs-app is the pod which utilizes the EFS PVC. For the most accurate logs, please substitute the pod name corresponding to the aforementioned node in the below steps for <driver_pod_name>.

Log collector script

The log collector script will collect

  • output of kubectl describe pod <driver_pod_name> -n kube-system
  • output of kubectl get pod <driver_pod_name> -o yaml -n kube-system
  • efs-utils logs from /var/log/amazon/efs
  • efs-utils state files from /var/run/efs
  • nfs mounts on filesystem

You can run the log collector as follows:

python3 log_collector.py --driver-pod-name <driver_pod_name>
# for example
python3 log_collector.py --driver-pod-name efs-csi-node-7g8k2

However, before running the log collector, you may want to enable debug logs for both the csi driver and efs-utils, which is the mounting utility used by the driver.

After enabling debug logging using the following sections and before collecting logs, make sure to attempt the mount again, so that new failure logs are acquired.

Enable debug logging for efs csi driver

Modify the deployment manifests to use the new logging level with the following command from deploy/kubernetes or charts directory (depending on your installation method):
find . -type f | xargs sed -i '/v=2/s//v=5/g'

You will need to redeploy the driver after running this command.

Enable debug logging for efs-utils

  • kubectl exec <driver_pod_name> -n kube-system -it /bin/bash
  • From inside the pod, sed -i '/stunnel_debug_enabled = false/s//stunnel_debug_enabled = true/g' /etc/amazon/efs/efs-utils.conf
  • From inside the pod, sed -i '/logging_level = INFO/s//logging_level = DEBUG/g' /etc/amazon/efs/efs-utils.conf

Advanced debugging (optional)

Capture traffic from the pod where failure is occurring:
tcpdump -W 30 -C 1000 -s 2000 -w nfs_pcap_$(date +%FT%T).pcap -i any -z gzip -Z root 'port 2049 or (src 127.0.0.1 and dst 127.0.0.1)'

Capture operating system log data from the pod where failure is occurring:
cat /var/log/message > kernel_debug.log

Posting to this repository's issues

We encourage users to post errors that you may run into to this repository's issues. Please include the appropriate logs from the above sources.