[AMBARI-8768] Ambari agent Heartbeat lost when df hangs (NFS gateway), also prevents proper re-initialization of agent upon restart

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.7.0
• Fix Version/s: 2.1.0
• Component/s: ambari-agent

Environment

HDP 2.1

• Assignee: Dmytro Sen
• Reporter: Hari Sekhon
• Created: 17/Dec/14 19:46
• Updated: 29/Sep/15 05:01
• Resolved: 20/May/15 16:25

Description

Ambari agent is succeptible to hanging when the 'df' command blocks. This causes loss of heartbeat and manageability. I've found this has happened with NFS gateway's HDFS mount point blocking when HDFS isn't available (we had set the NFS soft option on the mount point but then realized that wasn't a good idea as not everyone's processes and scripts will handle failure gracefully and retry properly).

When restarting the agent it also leaves the df process bound to point 8670 which requires manually killing that in order to get the ambari agent to restart and bind successfully, but even then you'll see a hang at this point after connecting to the 8440 ca and the agent never fully initializes so the heartbeat still never comes back.

The df command should be either in another thread non-blocking the main heartbeat and management functions or should have a timeout set on the command execution to prevent this issue.

Regards,

Hari Sekhon
http://www.linkedin.com/in/harisekhon

Issue Links

• links to: Review Board