The NEAR RPC node set up for #27 (neard 2.13.4 on frankie, tracking intents.near's shard) is reachable only on the mesh, as https://near.internal on the hanzalova proxy. It should also have a public name, alongside the <chain>.blackbeard.observer node endpoints that already exist for Quantus and Planck.
What
near.blackbeard.observer on the oolon edge, through the existing script/infra-setup.sh --role nodes path: a Cloudflare CNAME to the site indirection, a Let's Encrypt cert via DNS-01 with the Cloudflare credentials on oolon, and the vhost.
Upstream is https://near.internal, not frankie directly. The node moves to new hardware next week, and this way the move changes one internal vhost rather than two vhosts and a firewall. It also leaves frankie's firewall admitting only the hanzalova proxy and bob.
Throttling, tighter than the Quantus endpoint's, because the node is weak and the observer's own indexer shares it:
per client: near_rpc at 5 r/s, burst 20, and 8 concurrent connections, answering 429
a global ceiling across all clients, so many addresses together cannot starve the indexer (which reads via near.internal and is not subject to the public limits)
JSON-RPC only: POST /, plus GET /status and /health; small bodies; no WebSocket, since NEAR RPC has none
Refuse Hetzner and OVH hosting networks with the same $blackbeard_denied geo list the site vhost uses (#24), and give it its own JSON access log in the blackbeard_json format.
Not this
It does not fix the node's sync. State sync needs peers to connect back on TCP 24567 (P2P), which is a separate decision about the office WAN. Until then the public name serves a node that reports syncing: true.
Nothing about the intents.near-only tracking changes. Queries about accounts on other shards will fail on this node, publicly as they do privately.
## Why
The NEAR RPC node set up for #27 (neard 2.13.4 on frankie, tracking `intents.near`'s shard) is reachable only on the mesh, as `https://near.internal` on the hanzalova proxy. It should also have a public name, alongside the `<chain>.blackbeard.observer` node endpoints that already exist for Quantus and Planck.
## What
- `near.blackbeard.observer` on the oolon edge, through the existing `script/infra-setup.sh --role nodes` path: a Cloudflare CNAME to the site indirection, a Let's Encrypt cert via DNS-01 with the Cloudflare credentials on oolon, and the vhost.
- **Upstream is `https://near.internal`, not frankie directly.** The node moves to new hardware next week, and this way the move changes one internal vhost rather than two vhosts and a firewall. It also leaves frankie's firewall admitting only the hanzalova proxy and bob.
- **Throttling, tighter than the Quantus endpoint's**, because the node is weak and the observer's own indexer shares it:
- per client: `near_rpc` at 5 r/s, burst 20, and 8 concurrent connections, answering 429
- a **global ceiling** across all clients, so many addresses together cannot starve the indexer (which reads via `near.internal` and is not subject to the public limits)
- JSON-RPC only: `POST /`, plus `GET /status` and `/health`; small bodies; no WebSocket, since NEAR RPC has none
- **Refuse Hetzner and OVH hosting networks** with the same `$blackbeard_denied` geo list the site vhost uses (#24), and give it its own JSON access log in the `blackbeard_json` format.
## Not this
- **It does not fix the node's sync.** State sync needs peers to connect back on TCP 24567 (P2P), which is a separate decision about the office WAN. Until then the public name serves a node that reports `syncing: true`.
- Nothing about the `intents.near`-only tracking changes. Queries about accounts on other shards will fail on this node, publicly as they do privately.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Why
The NEAR RPC node set up for #27 (neard 2.13.4 on frankie, tracking
intents.near's shard) is reachable only on the mesh, ashttps://near.internalon the hanzalova proxy. It should also have a public name, alongside the<chain>.blackbeard.observernode endpoints that already exist for Quantus and Planck.What
near.blackbeard.observeron the oolon edge, through the existingscript/infra-setup.sh --role nodespath: a Cloudflare CNAME to the site indirection, a Let's Encrypt cert via DNS-01 with the Cloudflare credentials on oolon, and the vhost.https://near.internal, not frankie directly. The node moves to new hardware next week, and this way the move changes one internal vhost rather than two vhosts and a firewall. It also leaves frankie's firewall admitting only the hanzalova proxy and bob.near_rpcat 5 r/s, burst 20, and 8 concurrent connections, answering 429near.internaland is not subject to the public limits)POST /, plusGET /statusand/health; small bodies; no WebSocket, since NEAR RPC has none$blackbeard_deniedgeo list the site vhost uses (#24), and give it its own JSON access log in theblackbeard_jsonformat.Not this
syncing: true.intents.near-only tracking changes. Queries about accounts on other shards will fail on this node, publicly as they do privately.