OTA Update Fails Randomly Due to Missing stream-next File

Board Integrations
, ,
1

Hello,

I’m experiencing an issue with Mender OTA updates where roughly one out of three full rootfs updates fails.
The artifact is a full rootfs update, generated automatically by Yocto during the build process (the .mender artifact produced by the Yocto build).

The failure looks like this:

2025-11-21 13:20:29.232 +0000 UTC info: Running Mender client 5.0.3
2025-11-21 13:20:29.232 +0000 UTC info: Deployment with ID dba37c41-3b03-4f59-a183-fe58fb5d0170 started.
2025-11-21 13:20:29.233 +0000 UTC info: Sending status update to server
2025-11-21 13:20:29.397 +0000 UTC info: Installing artifact...
[...]
2025-11-21 13:21:02.924 +0000 UTC info: Update Module output (stderr): cat: can't open 'stream-next': No such file or directory
2025-11-21 13:21:02.926 +0000 UTC error: Broken pipe: Update Module started downloading, but did not finish
2025-11-21 13:21:02.927 +0000 UTC error: Operation canceled: GET <...>: HTTP request cancelled
[...]
2025-11-21 13:21:12.529 +0000 UTC error: No such file or directory: Cannot open /var/lib/mender/modules/v3/payloads/0000/tree/stream-next

The failure is always related to the missing stream-next file.
What I find confusing is that the exact same artifact sometimes installs successfully, and other times it fails with this error, without any change on the server or the device.

This makes me wonder whether the issue comes from:

  • a race condition in the Update Module,
  • a temporary network interruption,
  • something happening in Mender’s streaming logic,
  • or a problem with how the rootfs OTA artifact is processed.

If anyone has encountered similar behavior or has any insights into what might cause the intermittent stream-next errors, I would really appreciate your help.

Context:

  • Yocto Scarthgap
  • Verdin iMX8MP
  • Mender client 5.0.3
  • Update Module used for full rootfs OTA (automatically generated .mender artifacts)

Thanks in advance!
Stan

2

Hi @Stan,

We’ve never seen this error context so far, so it’s hard to tell. A first guess might be an unusual filesystem or storage setup. Can you please provide the output of the mount command, /the /etc/fstab and the /proc/cmdline files?

Greetz,
Josef

3

Hi @TheYoctoJester,

Thanks for your reply.

I checked all the points you mentioned (mount layout, /etc/fstab, and /proc/cmdline), and everything looked correct on the system side. The actual root cause turned out not to be related to the filesystem configuration itself, but rather to the size of the data partition used by Mender.

The /data partition on my setup was too small, and during some updates the available space was insufficient for Mender’s streaming mechanism to buffer the stream-next chunks correctly. This explains why the issue was intermittent: depending on the amount of residual data or temporary files at the time of the update, the client sometimes had just enough space to complete the streaming stage, and sometimes not. When the free space dropped below what was needed, Mender failed while trying to create or access the stream-next file, resulting in the broken pipe and missing file errors.

After increasing the size of the data partition, the problem disappeared completely and all full rootfs updates have been installing reliably.

Thanks again for pointing me in the right direction!

1 Like