You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
k8s pod name is sometimes wrong for the IP - looks like it's related to hostnetwork:true so it's not easy to exclude those events entirely from telemetry, but we can remove k8s metadata from those events.
To consider as an intermediate: don't add k8s metadata to these events.
…nt IP (#260)
## Which problem is this PR solving?
When processing events, the agent tries to lookup the source and
destination IPs in the list of cached k8s pods. However, because the
agent uses host network, the agent gets the same IP as the node and any
other pods who also use host networking which leads to inconsistent
results.
This PR updates event processing to not lookup pod attributes if the
source or destination IP matches the agent IP.
- Closes#244
## Short description of the changes
- Move cached client utils functions onto the cached client as member
funcs
- Update GetAttr* funcs to also take agent IP
- Return from getting attributes early if pod IP and agent IP match
- Add unit test to verify behaviour
## How to verify that this has the expected result
When processing an event when the source or dest IPs match the agent IP,
k8s attributes are no longer added.
---------
Co-authored-by: Jamie Danielson <[email protected]>
k8s pod name is sometimes wrong for the IP - looks like it's related to
hostnetwork:true
so it's not easy to exclude those events entirely from telemetry, but we can remove k8s metadata from those events.Originally posted by @JamieDanielson in #240 (comment)
The text was updated successfully, but these errors were encountered: