Check_d Endpoint Scan v3 for Celestia

Node Operators Validators
1

check_d Endpoint Scan v3 for Celestia: better monitoring, smarter failover

We’ve released v3 of Cumulo’s check_d Endpoint Scan for both Celestia mainnet and Mocha testnet.

This update significantly expands the amount of operational information exposed for each public validator endpoint and introduces a new API layer for automatically selecting the best available RPC or REST endpoint.

What is check_d?

check_d continuously tests public validator RPC and REST/API endpoints from multiple independent geographic locations.

Current probe regions: United States, Europe, Canada

Instead of relying on ICMP pings or synthetic availability checks, check_d performs real protocol requests and tracks:

  • actual response latency

  • block height

  • endpoint availability

  • historical reliability

The goal is to provide a practical view of how endpoints perform from the perspective of real users and applications.

16 Celestia Check-d 01
16 Celestia Check-d 011247×626 136 KB

Who is it useful for?

The tool is mainly intended for:

  • Validators comparing the availability and quality of their public infrastructure against the rest of the network.

  • dApp and indexer developers looking for the best-performing endpoints by region.

  • Wallets and infrastructure tools that need automatic failover without hardcoding a single RPC or REST endpoint.

Live scans

Celestia mainnet

RPC:
https://cumulo.pro/services/celestia/rpcscan.php

REST/API:
https://cumulo.pro/services/celestia/apiscan.php

Mocha testnet

RPC:
https://cumulo.pro/services/celestia_mocha/rpcscan.php

REST/API:
https://cumulo.pro/services/celestia_mocha/apiscan.php

What’s new in v3

1. More complete endpoint status

The scanner now exposes substantially more context than a simple up/down status.

Each endpoint includes:

  • Last check. The exact time that specific endpoint was last tested.

  • Last success. For currently failing endpoints, this shows when the endpoint last responded successfully, making it easier to distinguish transient failures from prolonged downtime.

  • Pruned vs Archival. Curated from our validator resources rather than inferred technically from the endpoint.

  • CORS support. Whether the endpoint returns the headers required for direct browser access, particularly useful for wallets and dApps.

  • TLS certificate status. Valid, expiring soon (less than 14 days), or broken. TLS is checked independently every hour so certificate issues remain visible even if the endpoint is unavailable for another reason.

  • Rate-limit hints. If the server exposes X-RateLimit-* or Retry-After headers, check_d displays them. This detection is entirely passive; no additional requests are generated to trigger rate limiting.

2. Smoothed latency to reduce flapping

In addition to the raw latency measured during each scan cycle, endpoints now expose an Exponentially Weighted Moving Average (EMA).

Raw latency can spike because of temporary network noise or a single slow request. The EMA provides a more stable representation of sustained endpoint performance.

This makes it a safer metric for automated endpoint selection.

3. Smart Endpoint Selector

This is the largest addition in v3.

We now expose a JSON API that applications can use to dynamically select the best available Celestia endpoint instead of hardcoding one.

RPC

Best current endpoint:

GET https://aggregate-rpcs.cumulo.com.es/best-rpc?chain=Celestia+mainnet

Full ranked list:

GET https://aggregate-rpcs.cumulo.com.es/rank-rpcs?chain=Celestia+mainnet

/best-rpc returns a single JSON object containing the current best candidate.

/rank-rpcs returns the full list of eligible endpoints, ordered by rank.

The same pattern is available for REST/API endpoints through:

aggregate-apis.cumulo.com.es

using:

  • /best-api

  • /rank-apis

Ranking logic

Only endpoints that are currently healthy are eligible.

Candidates are ranked by:

  1. historical reliability over the previous 7 days

  2. smoothed latency (EMA) as the tiebreaker

You can also request regional ranking:

&region=US

&region=EU

&region=CA

When a region is specified, the ranking uses latency measured from that specific region instead of the global average.

This can be useful when an application knows where most of its users or backend infrastructure are located.

We are also using the selector internally as an automatic fallback source for some of our own Celestia tools in case our primary endpoints become unavailable.

4. Redesigned monitoring panel

The frontend has also been reorganized so that the new operational data is visible directly from the main table.

Changes include:

  • compact endpoint capability/status badges under each validator

  • combined Latency / Reliability information

  • clearer visibility of endpoint health

  • the current Smart Endpoint Selector winner highlighted directly in the table with a :trophy: badge

This makes it possible to identify the current top-ranked endpoint without having to cross-reference the JSON API.

Documentation

The implementation and documentation are available here:

https://github.com/Cumulo-pro/Cumulo-Front-Chain/tree/main/check_End-Points

As with the rest of our Celestia tooling, the objective is to make validator infrastructure easier to inspect, compare and consume programmatically.

If you notice incorrect information for a particular endpoint, or there are additional metrics that would be useful for validators or builders, feel free to reply in this thread and we’ll take a look.

Cumulo

1 Like