Collect traffic levels #271
Conversation
orishavit
force-pushed
the
shavit/traffic-levels
branch
from
95d00a6 to
8b467c7
Compare
| sourceIdentity, err := r.resolveIPToIdentity(ctx, report.SrcIP) | ||
|
|
||
| if err != nil { | ||
| logrus.WithError(err).Errorf("Could not resolve source IP %s to identity", report.SrcIP) |
There was a problem hiding this comment.
use logrus.WithField for templated fields:
| logrus.WithError(err).Errorf("Could not resolve source IP %s to identity", report.SrcIP) | |
| logrus.WithError(err).WithFields("srcIP", report.SrcIP).Error("Could not resolve source IP to identity") |
| destinationIdentity, err := r.resolveIPToIdentity(ctx, report.DstIP) | ||
|
|
||
| if err != nil { | ||
| logrus.WithError(err).Errorf("Could not resolve destination IP %s to identity", report.DstIP) |
| @@ -314,6 +315,33 @@ func (r *Resolver) handleAzureOperationReport(ctx context.Context, operation mod | |||
| return nil | |||
| } | |||
|
|
|||
| func (r *Resolver) handleTrafficLevelReport(ctx context.Context, results model.TrafficLevelResults) error { | |||
There was a problem hiding this comment.
unfortunately, resolving source and dest pods just by trying to search for the IP in kubefinder is not that reliable, due to annoying things like IP reuse or host-network pods. Please follow the flow performed by handleTCPCaptureResult to correctly resolve the local and remote identities, taking all edge cases into consideration.
There was a problem hiding this comment.
The code in handleTCPCaptureResult also doesn't work well with traffic info, since it makes assumptions on the direction of traffic. Specifically, that traffic is always from a pod to a service, which is not true in our case. I'm not sure what's the best way to handle this generally, but I hesitate to modify the code used by handleTCPCaptureResult so to not break existing features.
Description
Aggregate data collected by the Otterize node-agent, using the traffic levels eBPF programs. This data includes the number of bytes, and number of connections between the services. The mapper then calculates a moving average, and reports it to the cloud.
Checklist