Network error on backup download

[Canapin (Coin-coin le Canapin)]

My co-admin told me he couldn’t download a backup from the received email because the download crashed at roughly 50%.

I tried and experienced the same. The archive download works well until it stops with “Failed - Network error”.
Trying to resume on Chrome returns “Failed - Unknown server error”. I had no issue the last time I downloaded backups from this server (months ago). (this is expected)

Any idea?

edit, more info:

Network error on backup download

I tried downloading a backup from another forum and experienced the same. Can someone try to replicate on their own instance? On both sites the download fails after 30 seconds. Moving this to bug because I’m quite sure this is not just me.

Repro steps:

1. Download a backup from the emailed link
2. The download should fail at some point

[MaxenceC (Maxence)]

Hi,

I’m Canapin’s co-admin

Trying to resume on Chrome returns “Failed - Unknown server error”. I had no issue the last time I downloaded backups from this server (months ago).

I’m wondering whether this is due to the one-time link. You can’t download a backup twice with a single link. So, when it fails once, resuming the download is probably cancelled by Discourse itself.

[Ethsim2 (Ethan)]

@Canapin

the download crashed at roughly 50%.

i think you need to change the way you’re downloading the backup

Backup upload not working

Press ⌘ + Space → type Terminal → Enter Run your command scp -C root@domain.com:/var/discourse/shared/standalone/backups/default/cate-s-community-2025-10-17-161933-v20251016143343.tar.gz . -C → enables compression root@domain.com: → your server’s login /var/discourse/…tar.gz → remote path . → save it in your current local directory (e.g. /Users/yourname)

[Canapin (Coin-coin le Canapin)]

Well, there sure are other ways to download backups, but I’d rather identify the cause and fix this particular issue

[Ethsim2 (Ethan)]

@MaxenceC

I’m wondering whether this is due to the one-time link. You can’t download a backup twice with a single link. So, when it fails once, resuming the download is probably cancelled by Discourse itself.

probably this, but this is not the only issue preventing full download of the backup via web browser

[Canapin (Coin-coin le Canapin)]

As an update, the forum was migrated to another server (for unrelated reasons), but the issue persists. The backup download (3.3 GB) always fails.

[Canapin (Coin-coin le Canapin)]

I tried downloading a backup from another forum and experienced the same.

Can someone try to replicate on their own instance? On both sites the download fails after 30 seconds.

Moving this to Contribute > Bug because I’m quite sure this is not just me.

[NJGF-Admin]

I have been experiencing this same issue for over a week. Fully updated twice in that time but the issue still persists. Self-hosted on IONOS. Backup size 1.5GB.

[Ethsim2 (Ethan)]

I’ve submitted a PR to address this by increasing the Nginx timeouts specifically for /admin/backups/, which should help prevent mid-stream “Failed – Network error” drops during larger downloads:

https://github.com/discourse/discourse/pull/36107

[NJGF-Admin]

Tried again just now. Still failing at 1GB

[sam (Sam Saffron)]

This is way too risky, it can open up risk for unintended denial of service.

We should be using sendfile there and nginx should handle it without any proxy needs.

[NJGF-Admin]

Fully updated 2025.12.0-latest, issue persists.

Edit: I can download the backups via WinSCP or something like that but it would be nice if it worked as advertised and the email link + browser download completes successfully.

AI is probably wrong but also advising to increase the: DISCOURSE_NGINX_PROXY_READ_TIMEOUT: 600 to resolve this issue.

[WaitroseCarpark (CARPARK)]

I’m getting this as well. 13GB backup size (my last full backup before offloading all media to S3 before anyone complains!)

[Ethsim2 (Ethan)]

@WaitroseCarpark

13GB backup size

wow! this is definitely the size to be doing over SSH rather than the web browser

[Canapin (Coin-coin le Canapin)]

Regardless of what should be the right way to do something, we’re just highlighting the fact that a useful built-in feature is currently broken

[Ethsim2 (Ethan)]

i was being sarcastic

@WaitroseCarpark

offloading all media to S3

the python script below won’t continue to work after enabling secure-media though

https://github.com/Ethsim12/Discourse-Image-Exporter-by-requests

[WaitroseCarpark (CARPARK)]

Might I suggest some helpful text telling admin file location of backup incase they want to download with a client?

Wouldn’t have found it without @Ethsim2 earlier post

[NJGF-Admin]

@WaitroseCarpark

file location of backup

/var/discourse/shared/standalone/backups/default/your_backup_filename.tar.gz

[WaitroseCarpark (CARPARK)]

Also this button is misleading - it would be better labelled as ‘Email Download Link’

[pfaffman (Jay Pfaffman)]

@WaitroseCarpark

Might I suggest some helpful text telling admin file location of backup incase they want to download with a client?

It doesn’t know. It can’t know. It depends on your docker configuration in your yml file. It’s usually in standalone, but it could be in web_only, but it could also be anywhere in the file system.

@WaitroseCarpark

Also this button is misleading - it would be better labelled as ‘Email Download Link’

While that’s technically true, I don’t remember anyone else expressing that concern in the past decade. Do you find it very misleading? You can change it on your own site if you think it’ll help.