1
0
mirror of https://github.com/openshift/openshift-docs.git synced 2026-02-05 12:46:18 +01:00
Files
openshift-docs/modules/network-observability-pktdrop-overview.adoc

44 lines
2.6 KiB
Plaintext

// Module included in the following assemblies:
//
// network_observability/observing-network-traffic.adoc
:_mod-docs-content-type: CONCEPT
[id="network-observability-pktdrop-overview_{context}"]
= Packet drop tracking
[role="_abstract"]
Monitor and analyze network packet loss by using eBPF-based packet drop tracking, which identifies drop locations, detects host or OVS-specific drop reasons, and provides dedicated graphical panels in the *Overview* view.
You can configure graphical representation of network flow records with packet loss in the *Overview* view. By employing eBPF tracepoint hooks, you can gain valuable insights into packet drops for TCP, UDP, SCTP, ICMPv4, and ICMPv6 protocols, which can result in the following actions:
* Identification: Pinpoint the exact locations and network paths where packet drops are occurring. Determine whether specific devices, interfaces, or routes are more prone to drops.
* Root cause analysis: Examine the data collected by the eBPF program to understand the causes of packet drops. For example, are they a result of congestion, buffer issues, or specific network events?
* Performance optimization: With a clearer picture of packet drops, you can take steps to optimize network performance, such as adjust buffer sizes, reconfigure routing paths, or implement Quality of Service (QoS) measures.
When packet drop tracking is enabled, you can see the following panels in the *Overview* by default:
* *Top X packet dropped state stacked with total*
* *Top X packet dropped cause stacked with total*
* *Top X average dropped packets rates*
* *Top X dropped packets rates stacked with total*
Other packet drop panels are available to add in *Manage panels*:
* *Top X average dropped bytes rates*
* *Top X dropped bytes rates stacked with total*
== Types of packet drops
Two kinds of packet drops are detected by network observability: host drops and OVS drops. Host drops are prefixed with `SKB_DROP` and OVS drops are prefixed with `OVS_DROP`. Dropped flows are shown in the side panel of the *Traffic flows* table along with a link to a description of each drop type. Examples of host drop reasons are as follows:
* `SKB_DROP_REASON_NO_SOCKET`: the packet dropped due to a missing socket.
* `SKB_DROP_REASON_TCP_CSUM`: the packet dropped due to a TCP checksum error.
Examples of OVS drops reasons are as follows:
* `OVS_DROP_LAST_ACTION`: OVS packets dropped due to an implicit drop action, for example due to a configured network policy.
* `OVS_DROP_IP_TTL`: OVS packets dropped due to an expired IP TTL.
See the _Additional resources_ of this section for more information about enabling and working with packet drop tracking.