Verdaccio as upstream to another Verdaccio in offline mode causes repeated 200 (0 bytes) → uplink timeouts → 500 errors

[ganeshthirumani-05]

Hi Verdaccio team 👋,

I’m trying to use Verdaccio as an upstream for another Verdaccio in an offline / air-gapped setup, but I’m running into repeated timeouts and 500 errors even though the upstream Verdaccio already has all required packages cached and works fine standalone in offline mode.

I’d like to understand whether this setup is supported, unsupported by design, or if I’m missing something in configuration.

npm client ↓ Verdaccio A (consumer registry) ↓ Verdaccio B (upstream registry – already cached, offline)

Verdaccio B

Fully populated with required packages

Works correctly in offline mode

Clients can install packages directly from it without issues

Verdaccio A

Uses Verdaccio B as an uplink

No direct access to npmjs.org

Intended to rely only on Verdaccio B

`
storage: /verdaccio/storage/data

auth:
htpasswd:
file: /verdaccio/auth/htpasswd
max_users: -1

uplinks:
patch-minor:
url: http://ip:9082
headers:
authorization: "Basic "
timeout: 30s
max_fails: 3
fail_timeout: 5m

patch-major:
url: http://ip:9081
headers:
authorization: "Basic "
timeout: 30s
max_fails: 3
fail_timeout: 5m

packages:
'@/':
proxy: [patch-minor, patch-major]
access: $authenticated
publish: $authenticated

'**':
proxy: [patch-minor, patch-major]
access: $authenticated
publish: $authenticated

server:
keepAliveTimeout: 0

middlewares:
audit:
enabled: false

log:
type: stdout
format: pretty
level: http
`

Observed Behavior

Verdaccio A receives a request for a package tarball

It returns HTTP 200 with bytes: 0/0 (no data)

Verdaccio then retries the same request against the uplink

After retries:

ESOCKETTIMEDOUT

500 internal server error

uplink marked as offline

Example logs:

`http <-- 200, user: null, req: 'GET /@jsonjoy.com/json-pack/-/json-pack-17.65.0.tgz', bytes: 0/0

info --- making request:
GET http://ip:9081/@jsonjoy.com/json-pack/-/json-pack-17.65.0.tgz

error --- unexpected error: ESOCKETTIMEDOUT

http <-- 500, req: 'GET /@jsonjoy.com/json-pack/-/json-pack-17.65.0.tgz',
error: internal server error

warn --- host ip:9081 is now offline
`

This repeats for many packages, especially scoped ones.

✅ What Works

Installing packages directly from Verdaccio B → works

Verdaccio B in offline mode → works

Verdaccio A + npmjs.org uplink → works

Issue happens only when Verdaccio → Verdaccio is used as uplink

❓ Questions

Is Verdaccio → Verdaccio chaining officially supported?

Is this behavior expected due to metadata / tarball streaming assumptions?

Are 200 (0 bytes) responses expected during metadata resolution?

Is there a recommended way to:

use one Verdaccio as a read-only upstream

in offline / air-gapped environments

If unsupported, is this a known limitation by design?

I came across the old issue below, which seems related but unresolved:

https://github.com/verdaccio/verdaccio/issues/1372?utm_

Goal

A stable setup where:

Verdaccio A consumes packages

Verdaccio B acts as a fully cached upstream

No external internet access is required

No repeated retries / uplink offline marking occurs

[mbtools]

I think this should work. Which version of Verdaccio are you using? What's the environment (docker, helm, ...)?

A middleware log (DEBUG=verdaccio:middleware*) should show you more details on the URL mapping. Can you add such log of A and B?

PS: You post is hard to read. Please fix the formatting of the coding blocks. Thx 🙏

[Tanimodori]

Maybe related: #5773