Hi team!
Here's a case that Ping-pong Effect Caused by Improper Networking Between ESX Hosts and Storage.
Symptom
The storage receives commands for switching LUNs continuously, which leads to ping-pong effect.
1. Multiple ESX hosts are connected to Huawei storage through improper networks.
Multiple RDM disks are taken over by multiple hosts simultaneously.
The multipathing software is NMP.
2. The ALUA function is enabled.
3. The path selection algorithm of NMP is set to MRU.
Fault Diagnosis
Multiple RDM disks are taken over by multiple ESX hosts.
The path selection algorithm, MRU, selects the path most recently used.
In the non-dual-switch network, some hosts are connected only from one controller.
Based on the previous conditions, ping-pong effect is caused.
Solution
1. Establish a vCenter cluster for multiple ESX hosts.
2. Modify the networking mode to standard dual-switch networking.
3. Set the path selection algorithm to fixed, indicating that I/Os are delivered from a fixed path.
This algorithm is recommended by VMware because it has a better compatibility.
0
