mirror of
https://github.com/Ride-The-Lightning/RTL.git
synced 2026-08-13 12:33:07 +02:00
Broaden release note to cover LND peers parity and getAlias hardening (#1629)
This commit is contained in:
parent
bd74132265
commit
e11899a051
1 changed files with 13 additions and 11 deletions
|
|
@ -70,18 +70,20 @@ this release should add its entry under the appropriate section below.
|
|||
(WCAG 2.4.3) that produced an inconsistent keyboard order; these were removed so focus follows
|
||||
natural DOM order across the LND, Core Lightning, Eclair and shared modals.
|
||||
|
||||
- **Core Lightning: bound alias resolution on the peers and route lookups to stop clnrest
|
||||
"Resource temporarily unavailable" errors** ([#1629](https://github.com/Ride-The-Lightning/RTL/pull/1629),
|
||||
- **Bound peer/route alias resolution to stop clnrest "Resource temporarily unavailable" errors**
|
||||
([#1629](https://github.com/Ride-The-Lightning/RTL/pull/1629),
|
||||
fixes [#1501](https://github.com/Ride-The-Lightning/RTL/issues/1501)).
|
||||
RTL resolves peer aliases by calling `listnodes` once per peer. A prior fix bounded this to 20
|
||||
concurrent calls (plus a cache) for the channel list, but the **peers list** and **route lookup**
|
||||
still fired an unbounded `Promise.all` — one request per peer at once — which overwhelms clnrest
|
||||
on nodes with many peers and fails with `Resource temporarily unavailable (os error 11)`
|
||||
(`EAGAIN`), leaving raw node IDs instead of aliases. Both paths now use the same 20-way
|
||||
concurrency limit. The limiter was also hardened to resolve immediately for an empty list (an
|
||||
empty peers/route set would previously never send a response), and the alias cache gained a
|
||||
6-hour TTL and a max size so aliases refresh without an RTL restart and the cache can't grow
|
||||
unbounded.
|
||||
RTL resolves peer aliases by calling `listnodes` (CLN) / `graph/node` (LND) once per peer. A
|
||||
prior fix bounded this to 20 concurrent calls (plus a cache) for the CLN channel list, but the
|
||||
Core Lightning **peers list** and **route lookup** — and the LND **peers list** — still fired an
|
||||
unbounded `Promise.all`, one request per peer at once. On Core Lightning nodes with many peers
|
||||
this overwhelms clnrest and fails with `Resource temporarily unavailable (os error 11)`
|
||||
(`EAGAIN`), leaving raw node IDs instead of aliases. All of these paths now use the same 20-way
|
||||
concurrency limit (Eclair already resolves aliases inline from a bulk nodes list, so it is
|
||||
unaffected). The CLN alias lookup was also made self-contained so aliases resolve regardless of
|
||||
which screen is opened first; the limiter now resolves immediately for an empty or non-positive
|
||||
input (which would previously never send a response); and the CLN alias cache gained a 6-hour TTL
|
||||
and a max size so aliases refresh without an RTL restart and the cache can't grow unbounded.
|
||||
|
||||
## Enhancements
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue