Node Network Traffic counters reporting very large values

[codekoala]

Describe the bug When viewing the px/nodes and px/node scripts, two of the nodes in a DigitalOcean Kubernetes cluster are reporting very large values for both sent and received network traffic at irregular intervals.

To Reproduce Steps to reproduce the behavior:

1. Go to a cluster in Pixie
2. Click on script px/nodes
3. Set start_time to -5h (any value works, but trends are easier to see with a larger window)
4. Click "Run"
5. Scroll down to "Sent Network Traffic" and "Received Network Traffic"

Expected behavior I expect to see a graph indicating steady amounts of traffic, with occasional spikes.

Screenshots

Using a 10-minute window to showcase the irregular intervals:

Tabular view of data from screenshot above:

Graph of node traffic from DigitalOcean's console:

Logs Please attach the logs by running the following command:

./px collect-logs

pixie_logs_20210601074131.zip

App information (please complete the following information):

• Pixie version: 0.5.12+Distribution.a945b68.20210525150228.1
• K8s cluster version: 1.20.2-do.0

Additional context There's not really anything running on this cluster that should explain any kind of spikes in network traffic. It is a disposable cluster, and normally I would have deleted it and started over by now. I'm trying to leave it alone just in case it's useful for debugging this behavior.