PowerFlex: MDM Flaps error TX driver issue detected

 MDM keeps flapping between DEGRADED and NORMAL states. Other services (SDS, SDC) running on the same host can also be affected.

Scenario

The issue can happen during normal operations.

Symptoms

Affected host /var/log/messages are showing:
 

Mar 8 06:32:00 sds4-4 kernel: i40e 0000:0a:00.0: TX driver issue detected, PF reset issued
Mar 8 06:32:01 sds4-4 kernel: i40e 0000:0a:00.0: i40e_ptp_init: added PHC on ens3f0
Mar 8 06:32:01 sds4-4 kernel: i40e 0000:0a:00.0: enabling bridge mode: VEPA
Mar 8 06:32:01 sds4-4 kernel: i40e 0000:0a:00.0 ens3f0: NIC Link is Up 10 Gbps Full Duplex, Flow Control: None
Mar 8 06:32:01 sds4-4 kernel: i40e 0000:0a:00.0: TX driver issue detected, PF reset issued
Mar 8 06:32:02 sds4-4 kernel: i40e 0000:0a:00.0: i40e_ptp_init: added PHC on ens3f0
Mar 8 06:32:02 sds4-4 kernel: i40e 0000:0a:00.0: enabling bridge mode: VEPA
Mar 8 06:32:02 sds4-4 kernel: i40e 0000:0a:00.0 ens3f0: NIC Link is Up 10 Gbps Full Duplex, Flow Control: None
Mar 8 06:32:02 sds4-4 kernel: i40e 0000:0a:00.0: TX driver issue detected, PF reset issued
Mar 8 06:32:03 sds4-4 kernel: i40e 0000:0a:00.0: i40e_ptp_init: added PHC on ens3f0
Mar 8 06:32:03 sds4-4 kernel: i40e 0000:0a:00.0: enabling bridge mode: VEPA
Mar 8 06:32:03 sds4-4 kernel: i40e 0000:0a:00.0 ens3f0: NIC Link is Up 10 Gbps Full Duplex, Flow Control: None

Impact

The stability of ScaleIO services running on the host can be affected.

The problem caused by the software issue with the NIC driver built into the Linux kernel

Disable both the TSO and GSO offloading functions:

# ethtool -K ethX tso off gso off

 For a permanent fix, upgrade the kernel to kernel-3.10.0-514.17.1.el7 or later (see: i40e: TX driver issue detected, PF reset issued - Red Hat Customer Portal