Why does non-mandatory FPolicy have latency when EAGAIN is logged?
Applies to
- ONTAP 9
- CIFS
- NFS
- FPolicy with mandatory screening set to false
- FPolicy in asynchronous mode
Answer
- FPolicy is guaranteed
- The FPolicy server must acknowledge requests
- The FPolicy server is not required to complete acknowledged requests
- ONTAP utilizes the TCP send-buffer to queue requests
- In asynchronous mode, file operations can be processed by ONTAP when a request is queued
- The send-buffer is not infinite, providing a limit to the number of queued requests
- The send-buffer can be increased from the default setting, if necessary
- An FPolicy server that is slow to respond may cause the send-buffer to reach capacity
- Existing requests are sent to the back of the queue via EAGAIN
- New requests are delayed until TCP acknowledgements empty sufficient space in the send-buffer
- Latency is experienced while the send-buffer is full