Release Notes

v6.2 EE Release Notes

Created：2024-11-05 Last Modified：2024-11-05

This document was translated by ChatGPT

#1. Applications, Network, Metrics, Events

AutoTracing
- Support zero-instrumentation distributed tracing for Golang applications
- Add file IO event performance data and associate it with call logs
Universal Service Map
- Support zero-instrumentation automatic display of process-level universal application topology
- Support calculating the real access relationships before and after NAT using the TOA (TCP Option Address) mechanism
- Pre-aggregate OpenTelemetry Span data into service and path metrics
- Correct path data direction using the operating system's socket information
- Add direction score metric; the higher the score, the more accurate the client and server direction. A score of 255 indicates the direction is definitely correct.
- Support displaying application topology grouped by Custom Tag (e.g., k8s.label.xxx)
Private protocol parsing, business field parsing
- Introduce WebAssembly Plugin mechanism to support adding private protocol parsing and enhancing standard protocol field parsing capabilities, see the specific documentation here
New continuous profiling feature
- Support collection and display of Java and Golang Native Profiles
Network NAT tracing, flow logs, traffic download
- No rate limit for flow logs associated with PCAP policies
- Support viewing flow logs for specified PCAP policies
- Set the status of non-TCP flow logs to normal
- Reduce the minimum time span for PCAP downloads to 1s
- Support direct traffic download based on PCAP policies
Cloud network traffic distribution
- Support traffic distribution using TCP tunnels
- Display the size of traffic matched by distribution policies separately by distribution point
- Support pre-configuring workloads and container services not yet created in distribution policies
Usability improvements
- Data association
  - Support switching grouping conditions in the right slide-out frame metrics tab for easier drill-down
- Page operations
  - Expand the field types supported by template variables
  - Add quick filter functionality on the left side of table pages
  - Support search filtering in the advanced configuration dropdown for metrics
  - Improve the usability of the table column selection
  - Support direct PCAP file download from the TCP sequence diagram page
  - Support configuring metric templates for quick selection of common metric sets in Panels
  - Enhance control over the display of tips in Panels
  - Support quick copy of tags in the right slide-out page and flow log call log detail page, and support quick filtering
  - Optimize the filtering logic in the traffic relationship tab of the right slide-out frame to avoid ambiguity
  - Support setting historical search conditions as the default content loaded on the page
  - Change the full-screen display of all Panels to pop-up display
  - Support directly modifying path options in the search box
  - Optimize the operation and display of the distributed tracing flame graph, and support filtering to display APP/SYS/NET Span
- New pages
  - Add a topology view to distributed tracing
  - Add a Service List page to summarize service-level metric data from OTel, eBPF, and Packet signal sources
  - Add an Endpoint monitoring page, supporting association expansion from the service list page
- Streamlined pages
  - Merge path, flow log/call log pages
- Page display
  - Optimize the display of resources corresponding to the network path collection NIC; show as uncollected when no corresponding resource is found
  - Optimize the display of knowledge graph information in the right slide-out page
  - Use waterfall topology for NAT tracing page and add physical topology display
  - Optimize the display of network paths in the right slide-out page, supporting default display of physical network paths
  - Do not display empty value tags on flow log, call log, and other pages
  - Support adding description information for historical search conditions
  - Optimize the text display of search box fields, providing more helpful information
  - Display the values and description information of enum fields in data tables
  - Preemptively gray out the jump button when the target page has no data
  - Adjust and optimize page layout for more reasonable space utilization and increased information display
  - Use auto_instance, auto_service to replace the hard-to-understand resource_gl0/gl1/gl2

#2. Dashboards, Alerts, Reports

Optimize the display of trigger values in alert events, showing the current value instead of the trigger value
Review and refine the wording of all system alerts
Packet loss environment threshold alert policies for all collectors and data nodes

#3. Resources

AutoTagging
- Support synchronizing process information on container nodes and cloud servers, and inject it as Universal Tag into all data, supporting filtering and grouping
- Support adding custom metadata for processes, cloud servers, and K8s Namespaces
- Support automatic synchronization of K8s cluster information under AWS and Alibaba Cloud accounts
- Support synchronizing Baidu Cloud Cloud Smart Network (CSN) resource information
- Add support for K8s Service in the Custom Tag field k8s.label
- When a K8s Pod serves as the backend for multiple K8s Services, associate only the service with the smallest dictionary order in all tags
Usability
- Support batch input of load balancer and its listener information
- Support batch input of Gateway type host and its NIC
- Support unified setting of additional docking route interfaces for all managed K8s clusters under a public cloud account
Compatibility
- Adapt resource information synchronization for K8s 1.18, 1.20
- Add two configuration items for Huawei public cloud: domain name and IAM authorization address, to adapt to HCSO scenarios

#4. System

API
- Querier adds support for PromQL
  - Support querying DeepFlow native metrics via PromQL
  - Support querying Prometheus RemoteWrite metrics via PromQL
  - When querying Prometheus native metrics, support using tags automatically injected by DeepFlow AutoTagging
- Ingester adds support for OTLP Export
  - Support exporting SYS/NET Span data (l7_flow_log) to otel-collector, see the documentation for output fields here, FR-014-Tencent (opens new window).
- Querier SQL API
  - Provide Trace Completion API tracing-completion-by-external-app-spans: the caller passes in a set of APP Spans (no need to store them in DeepFlow) to get their upstream and downstream SYS/NET Spans, enhancing traditional APM with infrastructure and non-instrumented service tracing capabilities, see the specific documentation here
  - Add support for the AS keyword and return the original field name before AS in the results
  - When getting optional values for enum type Tag fields, return the description information corresponding to the values
  - Add SLIMIT parameter to limit the number of Series in the returned results
  - Unify the Category of custom type Tags (k8s.label/cloud.tag/os.app) as map_item
  - Add three Universal Tags: tap_port_host, tap_port_chost, tap_port_pod_node, representing the host, cloud server, and container node to which the collection NIC belongs, respectively
Grafana
- Prometheus Dashboard can use DeepFlow as a data source without modification
- In the Query Editor, support referencing the $__interval variable for the Math operator
- Optimize Enum type Variable to avoid expanding all candidate values in SQL when selecting All
- Add Grafana backend plugin module to support standard Grafana alert policy configuration
- Support referencing another Variable when defining a Variable
- Display the corresponding SQL statement in real-time when editing Query Editor content
eBPF
- When the application process only acts as a client, avoid eBPF associating all requests to the same Trace
- Simplify protocol recognition logic in eBPF code
Compatibility
- Support running the collector in Linux Kernel environments below 3.0
- Support specifying the Hostname of the environment where the collector is located, suitable for environments with hostname conflicts
- Provide two types of deepflow-agent binary packages: dynamic linking and static linking; the former depends on the glibc dynamic link library, while the latter has significant malloc/free lock contention under multithreading
- eBPF programs support automatic adaptation of kernel offsets using BTF files
Controller and data node management
- Support ClickHouse nodes greater than the number of deepflow-server replicas
- Support specifying domain name form of controller or ingester address for deepflow-agent
- Support configuring data storage duration in hourly granularity
- Change the configuration API for data duration to asynchronous to avoid timeout
- Add self-monitoring for the report service
Collector management
- Support specifying Group ID when creating collector groups (agent-group)
- Support running deepflow-agent as a regular process (instead of a Pod) on K8s Node
- When the ctrl_ip or ctrl_mac of the agent's operating environment changes, support automatically updating the agent's corresponding information to avoid disconnection
- Remote upgrade of deepflow-agent on cloud servers can be fully completed through deepflow-ctl, without manually mounting hostPath for deepflow-server
- Support configuring the time interval for Agent list k8s resources to reduce pressure on apiserver
- When only network monitoring authorization is available, add support for viewing HTTP and DNS under the application menu
- Automatically delete Agents that have been disconnected for more than vtap_auto_delete_interval, with a default deletion time of more than 1 hour of disconnection
Debug
- Lower the collection interval of all DeepFlow self-monitoring metrics to 10s
- Embed eBPF data source information (syscall, go-tls, go-http2, openssl, io-event) in the tap_port field to enhance debugging capabilities
- deepflow-server supports monitoring itself through continuous profiling
A series of performance optimizations for various software components