NFS UDP traffic intermittently drops after upgrade to ONTAP 9.2 or newer
Applies to
- ONTAP 9.2 and later
- NFSv3 over UDP
- All non TCP traffic
- ICMP
Issue
- Mostly the issue is found after ONTAP is upgraded to 9.2 or later, the filer does not reply to every ICMP (ping) request even if it can be found in a packet trace.
- This is occurring on every interface on the cluster and even on e0M interface.
- A strange behavior could be related to multiple LIFs on the same port not responding at different times.
- This issue is affecting the NFS data transfer also if UDP is used as a transport layer causing a huge delay
This can be confirmed by following simple test:
Situation1# Mounted with UDP
[root@centos-04 /]# mount -t nfs -o vers=3,proto=udp 10.128.242.45:/vol1_svm2 /mnt/vol1_svm2_ontap93
[root@centos-04 /]# cd /mnt/vol1_svm2_ontap93
[root@centos-04 vol1_svm2_ontap93]# dd if=/dev/zero of=./test bs=4096 count=10000
10000 0 records in
10000 0 records out
40960000 bytes (41 MB) copied, 37.0209 s, 1.1 MB/s <<<
Situation2# Mounted with TCP
[root@centos-04 vol1_svm2_ontap93]# cd /
[root@centos-04 /]# umount /mnt/vol1_svm2_ontap93
[root@centos-04 /]# mount -t nfs -o vers=3,proto=tcp 10.128.242.45:/vol1_svm2 /mnt/vol1_svm2_ontap93
[root@centos-04 /]# cd /mnt/vol1_svm2_ontap93
[root@centos-04 vol1_svm2_ontap93]# dd if=/dev/zero of=./test bs=4096 count=10000
10000 0 records in
10000 0 records out
40960000 bytes (41 MB) copied, 0.240958 s, 170 MB/s <<<