NFS file copy operations are slow due to packet loss and TCP retransmissions
Applies to
- ONTAP 9
- NFS
- Linux NFS clients
Issue
- Users experience significant performance degradation when accessing data on specific NFS shares.
-
Copying files from the NFS export to local directories taks significantly longer than expected.
- A 1 GB file copy operation that normally completed within 30 seconds to 2 minutes takes much longer e.g 30 minutes or 1.5 hours.
- Basic filesystem operations such as
lsexperience delays of 10 to 15 seconds, whereas they previously completed almost instantly. - Wireshark filter
tcp.analysis.fast_retransmission or tcp.analysis.retransmission or tcp.analysis.out_of_order or tcp.analysis.duplicate_ackshows repeated retransmissions of traffic -
Client-side connectivity tests line ping show packet
-
100 packets transmitted, 83 received, 17% packet loss
-
- ONTAP performance statistics indicated normal storage latency and CPU utilization during the affected timeframe, with no signs of a storage-side bottleneck.
