Latency between VM's is observed on destination Host when a VM is vMotioned or powered on/off.
book
Article ID: 336817
calendar_today
Updated On:
Products
VMware NSX
Issue/Introduction
This article explains what it is that is causing these latency spikes as well as how to mitigate/work around this problem.
Symptoms: When the customer tries to power on/off or migrate a VM (all flavors) from host to host they would observe a spike in the latency of up to 1,200 ms with ping drops between VM to VM on the destination host, which then settles back to about 1-2ms. This happens with all VM's on a host (including those on a DFW exclusion list), and with VM's on a standard switch. This spike lasts for about 20 seconds.
Environment
VMware NSX-T Data Center 3.x VMware NSX-T Data Center
Cause
Found in: NSX-T 3.0.2
Configuration pushes from the control plane causes significant delays in vsip packet processing. This is also seen during a vMotion import, which triggers the config pushes. This occurs with a large DFW ruleset and addrset config. The impact increases as more VMs are present on a host.
Resolution
Its fixed in NSX-T 3.1.3 GA Build 18328989 and later releases.
Workaround: The impact on the system of a configuration push is minimized with the reduction of rulesets and address sets. Effective use the AppliedTo field is necessary to lessen the impact of the config push. Otherwise, users need to stop configuration changes and vMotion activities to avoid the observable latencies in vsip's packet processing.