How to Collect Endpoint Details from yb-master and yb-tserver – Yugabyte

Summary

YugabyteDB consists of two main server components:

yb-master: Manages cluster metadata and tablet assignment.
yb-tserver: Handles user data and client requests.

Collecting endpoint details is important for troubleshooting, monitoring, and understanding the current state of your cluster. This guide provides step-by-step instructions to collect endpoint details from both yb-master and yb-tserver, including the most common and useful endpoints.

1. Collecting Endpoint Details via the Web UI

Both yb-master and yb-tserver expose a web UI for monitoring and diagnostics.

yb-master Web UI

Default URL:
http://<master_ip>:7000/

Steps:

Open a web browser.
Enter the URL: http://<master_ip>:7000/
The main dashboard displays master server details, tablet servers, and tablets.
Useful subpages:
- /memz (e.g., http://<master_ip>:7000/memz) — memory tracker details
- /prometheus-metrics (e.g., http://<master_ip>:7000/prometheus-metrics) — Prometheus metrics
- /tablet-servers — list of tablet servers
- /tablets — list of tablets

yb-tserver Web UI

Default URL:
http://<tserver_ip>:9000/

Steps:

Open a web browser.
Enter the URL: http://<tserver_ip>:9000/
The main dashboard provides tserver details, tablets, and performance metrics.
Useful subpages:
- /rpcz — RPC statistics
- /metrics — server metrics
- /mem-trackers — memory usage
- /threadz?group=all — thread information
- /pprof/contention — contention profiling

2. Collecting Endpoint Details Using `curl` (Command Line)

You can collect the same details programmatically from each node using curl by replacing <master_ip> or <tserver_ip> as appropriate.

yb-tserver Common Endpoints

Replace <tserver_ip> with your tserver node's IP or hostname:

# RPC statistics
curl -s "http://<tserver_ip>:9000/rpcz" > <tserver_ip>-rpcz.txt

# Metrics
curl -s "http://<tserver_ip>:9000/metrics" > <tserver_ip>-metrics.txt

# Memory trackers
curl -s "http://<tserver_ip>:9000/mem-trackers" > <tserver_ip>-mem-trackers.txt

# Thread details
curl -s "http://<tserver_ip>:9000/threadz?group=all" > <tserver_ip>-threadz.txt

# Contention profiling
curl -s "http://<tserver_ip>:9000/pprof/contention" > <tserver_ip>-contention.txt

yb-master Common Endpoints

Replace <master_ip> with your master node's IP or hostname:

# Memory tracker details (HTML output)
curl -s "http://<master_ip>:7000/memz" -o <master_ip>-memz.html

# Prometheus metrics (HTML output)
curl -s "http://<master_ip>:7000/prometheus-metrics" -o <master_ip>-prometheus-metrics.html

Tip: You can loop over multiple nodes to collect data from all servers in your cluster.

Summary Table of Useful Endpoints

Component	Endpoint	Description	Example URL
yb-tserver	`/rpcz`	RPC statistics	http://<tserver_ip>:9000/rpcz
yb-tserver	`/metrics`	Metrics	http://<tserver_ip>:9000/metrics
yb-tserver	`/mem-trackers`	Memory usage	http://<tserver_ip>:9000/mem-trackers
yb-tserver	`/threadz?group=all`	Thread info	http://<tserver_ip>:9000/threadz?group=all
yb-tserver	`/pprof/contention`	Contention profile	http://<tserver_ip>:9000/pprof/contention
yb-master	`/memz`	Memory tracker (HTML)	http://<master_ip>:7000/memz
yb-master	`/prometheus-metrics`	Prometheus metrics	http://<master_ip>:7000/prometheus-metrics
yb-master	`/tablet-servers`	Tablet servers list	http://<master_ip>:7000/tablet-servers
yb-master	`/tablets`	Tablets list	http://<master_ip>:7000/tablets

Tips for Support Collection

Save the output files with clear names indicating the node and endpoint.
Collect from all master and tserver nodes for completeness.
If automating, consider a shell script to loop over all hosts and endpoints.

Recommendation:
Use the UI for quick inspection and interactive troubleshooting. Use curl (command-line) for systematic collection, automation, and sharing with support.

Reference: SUPPORT-758