Warning: Connection Reset Issue in [4.9.0] which is n8n 2.10.2

[umnz]

Had to restore from backup this morning after being updated to 4.9.0.

At the time of writing, the latest update mentioned is 4.8.2 in this thread: https://forum.cloudron.io/post/120803

One of our workflows takes attachment binaries and uploads them via HTTP Requests. For some reason after updating to 4.9.0, these requests were getting ERRCONRESET and other associated connection errors when trying to remotely upload their files.

Other web requests, web hooks etc all worked fine.

Restoring a backup to 4.8.2 resolved the issue.

I can't see anyone else mentioning it on the n8n GitHub yet, but just a warning to other Cloudron users not to update (or to test this update on your workflows) before updating.

[paramedicspecialist]

I noticed a similar issue today after updating to 4.9.0 with one of our n8n workflows that posts binary files to StirlingPDF through the HTTP Request node.

The application logs for n8n show the following error:

Mar 04 13:33:02 Unable to calculate form data length
Mar 04 13:33:02 data should be a string, Buffer or Uint8Array

[joseph]

Not seeing anything in the changelog - https://github.com/n8n-io/n8n/releases . Maybe fixed in 2.10.3 ? But nothing related to this.

[joseph]

@paramedicspecialist I think yours is related to https://github.com/n8n-io/n8n/pull/25807 . The error messages match.

[umnz]

Please note this error persists into the latest version of n8n.

Specifically, I have a sub workflow that is executed that uses an HTTP request to upload files to a third party service. There are a lot of changes between 2.9.4 (working) and 2.10.2 (breaks). I'll continue to investigate and report back if I find anything.

https://github.com/n8n-io/n8n/issues/26748

[umnz]

@girish @Joseph n8n team are claiming the issue could be with changes made by Cloudron between 2.9.4 and above. Can you guys look into this?

There are multiple people claiming that sending binaries via multi-part-forms breaks after upgrading from 2.9.4 - but I don't know if they're using Cloudron.

Here's an example of someone getting the same error sending a multi-part form using the Open AI file upload:
https://github.com/n8n-io/n8n/issues/26777

Other references to connection issues when uploading files:
https://github.com/n8n-io/n8n/issues/26746
https://github.com/n8n-io/n8n/issues/26606
https://github.com/n8n-io/n8n/issues/25567

[zigmasdirigeant]

Same issue here. Workflow downloads audio from some of our tools and uploads to Azure OpenAI Whisper via multipart form. Azure logs confirm requestLength: 0 on every request. Container logs show the same "Unable to calculate form data length / data should be a string, Buffer, or Uint8Array" errors. Can Cloudron roll back to 4.8.2 or push a patched version? OR what's the plan here?
This broke 5 different workflows and things inside my whole enviroment, not critical, but data has been lost in this.

[joseph]

Just to update: we can reproduce this and are debugging this .

[james]

Hello @n8n users

@brutalbirdie debugged this issue and found the issue.
https://github.com/n8n-io/n8n/issues/26748#issuecomment-4031286237

We will update the Cloudron N8N app soon to fix this issue.

[umnz]

@James @brutalbirdie @Joseph

Much appreciated team.

[girish]

@umnz @paramedicspecialist I pushed a new package that does not use npm anymore and uses the upstream docker image. Can you please check?

[umnz]

@girish Sorry to report, but that update didn't fix the issue, in fact, it completely broke the n8n install.

Some sample errors from the logs before I had to restore from backup:

2026-03-12T14:05:51Z Activation of workflow "My Workflow 1" (FMTzz3R99WCvdr2P) did fail with error: "libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory" | retry in 32 seconds
2026-03-12T14:05:51Z libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory

That failure happened x every active/published workflow (40+). So I got a tonne of error notifications sent to Teams after the container started.

Is there a way to choose which update we want to jump to? Every release from 9.4.2 is broken so we are updating multiple times 0>1>2>3>4>Current, then reverting to 0 when it doesn't work.

[girish]

@umnz oh, that's not good. I will revoke the release for now.

[Lars.Emil]

Hi @staff,

I am experiencing a systemic failure in n8n version 2.11.3 (Cloudron package 4.9.x) where most workflows fail to activate or execute with the following error:

libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory

This appears to be a library mismatch introduced in a recent update. The error triggers specifically on nodes that utilize the JS sandbox (Merge nodes, Code nodes, etc.).

Technical findings:
I ran ldd on the isolated-vm binary inside the container, and it confirms that the module is compiled against musl (Alpine), while the Cloudron environment is glibc based (Ubuntu/Debian):

root@...:/app/code# ldd /app/code/node_modules/isolated-vm/out/isolated_vm.node
    linux-vdso.so.1 (0x00007ffca68bd000)
    libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x000071f4e608e000)
    libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x000071f4e6060000)
    libc.musl-x86_64.so.1 => not found
    libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x000071f4e5f77000)
    libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x000071f4e5d65000)
    /lib64/ld-linux-x86-64.so.2 (0x000071f4e6421000)

Observed behavior:

Workflow activation fails on startup for any workflow containing a Merge or Code node.

The Python task runner also fails to start in internal mode (virtual environment missing).

Setting N8N_BLOCK_JS_SANDBOX=true does not resolve the activation errors.

It seems the latest package might have been built/bundled in an Alpine environment by mistake, or a dependency has leaked a musl requirement into the glibc image.

Could you please look into recompiling the dependencies for the current base image?

[girish]

Thing is the Dockerfile is installing musl, so I am not sure what is going wrong here. Hmm

[girish]

@umnz so it doesn't even start up in your case or do you see this during a specific workflow? It installs and runs fine here

[umnz]

@girish I can access the application (it comes up like normal), and some workflows are working (for example, error workflow that calls teams), but a large majority the workflows don't come up properly and show that error on boot. Most of those have scheduled starting triggers or web hooks. I didn't have time to try republishing them because I wasn't expecting it to fail this way (which is a much bigger failure than the form binary upload issue that only affects 5-6 of our workflows). I've got the logs, I'll sanitize and post here shortly.

[girish]

@umnz I found the problem. The location of musl is different in alpine and ubuntu. Making a new package.

Our tests are not comprehensive enough to detect workflow errors. They only test if n8n comes up and one can login etc. Have to improve this.

[umnz]

Here's the logs from when I stopped and started the instance. I truncated the logs so at places where you see 'etc etc', just imagine another 20+ Workflows listed with the same entry or error.

2026-03-12T14:05:01Z Stopping n8n...
2026-03-12T14:05:01Z [runnner:js] Received SIGTERM signal, shutting down...
2026-03-12T14:05:01Z [runnner:js] Task runner stopped
2026-03-12T14:05:02Z 2026-03-12 14:05:02,724 WARN received SIGTERM indicating exit request
2026-03-12T14:05:02Z 2026-03-12 14:05:02,728 INFO waiting for redis, redis-service to die
2026-03-12T14:05:02Z 2026-03-12 14:05:02,733 WARN stopped: redis-service (terminated by SIGTERM)
2026-03-12T14:05:02Z 2026-03-12 14:05:02,840 INFO stopped: redis (exit status 0)
2026-03-12T14:05:02Z 41:M 12 Mar 2026 14:05:02.835 * Saving the final RDB snapshot before exiting.
2026-03-12T14:05:02Z 41:M 12 Mar 2026 14:05:02.835 * User requested shutdown...
2026-03-12T14:05:02Z 41:M 12 Mar 2026 14:05:02.837 * DB saved on disk
2026-03-12T14:05:02Z 41:M 12 Mar 2026 14:05:02.837 * Removing the pid file.
2026-03-12T14:05:02Z 41:M 12 Mar 2026 14:05:02.838 # Redis is now ready to exit, bye bye...
2026-03-12T14:05:02Z 41:signal-handler (1773324302) Received SIGTERM scheduling shutdown...
2026-03-12T14:05:03.402Z box:tasks updating task 1873 with: {"percent":100,"message":"Done"}
2026-03-12T14:05:03.409Z box:tasks setCompleted - 1873: {"result":null,"error":null,"percent":100}
2026-03-12T14:05:03.409Z box:tasks updating task 1873 with: {"completed":true,"result":null,"error":null,"percent":100}
2026-03-12T14:05:03.411Z Exiting with code 0
2026-03-12T14:05:03.411Z box:taskworker Task took 2.087 seconds
2026-03-12T14:05:08.974Z box:taskworker Starting task 1874. Logs are at /home/yellowtent/platformdata/logs/xyz/apptask.log
2026-03-12T14:05:09.017Z box:taskworker Running task of type app
2026-03-12T14:05:09.025Z box:apptask run: startTask installationState: pending_start runState: running
2026-03-12T14:05:09.025Z box:tasks updating task 1874 with: {"percent":10,"message":"Starting app services"}
2026-03-12T14:05:09.226Z box:tasks updating task 1874 with: {"percent":35,"message":"Starting container"}
2026-03-12T14:05:09.423Z box:tasks updating task 1874 with: {"percent":80,"message":"Configuring reverse proxy"}
2026-03-12T14:05:09.430Z box:shell reverseproxy: openssl x509 -noout -subject -issuer
2026-03-12T14:05:09.506Z box:reverseproxy providerMatches: subject=CN = workflow.domain.com domain=workflow.domain.com issuer=C = US, O = Let's Encrypt, CN = E8 wildcard=false/false prod=true/true issuerMismatch=false wildcardMismatch=false match=true
2026-03-12T14:05:09.506Z box:shell reverseproxy: openssl x509 -startdate -enddate -subject -noout
2026-03-12T14:05:09.578Z box:reverseproxy expiryDate: subject=CN = workflow.domain.com notBefore=Feb 23 11:12:14 2026 GMT notAfter=May 24 11:12:13 2026 GMT daysLeft=72.87990071759259
2026-03-12T14:05:09.579Z box:reverseproxy ensureCertificate: workflow.domain.com acme cert exists and is up to date
2026-03-12T14:05:09.579Z box:reverseproxy needsRenewal: false. force: false
2026-03-12T14:05:09.589Z box:reverseproxy writeAppLocationNginxConfig: writing config for "workflow.domain.com" to /home/yellowtent/platformdata/nginx/applications/xyz/workflow.domain.com.conf with options {"sourceDir":"/home/yellowtent/box","vhost":"workflow.domain.com","hasIPv6":true,"ip":"{redacted}","port":5678,"endpoint":"app","redirectTo":null,"certFilePath":"/home/yellowtent/platformdata/nginx/cert/workflow.domain.com.cert","keyFilePath":"/home/yellowtent/platformdata/nginx/cert/workflow.domain.com.key","robotsTxtQuoted":"\"# Disable search engine indexing\\n\\nUser-agent: *\\nDisallow: /\"","cspQuoted":null,"hideHeaders":[],"proxyAuth":{"enabled":false,"id":"xyz","location":"/"},"upstreamUri":"","hstsPreload":false}
2026-03-12T14:05:09.590Z box:shell reverseproxy: /usr/bin/sudo --non-interactive /home/yellowtent/box/src/scripts/restartservice.sh nginx
2026-03-12T14:05:09.649Z box:tasks updating task 1874 with: {"percent":100,"message":"Done"}
2026-03-12T14:05:09.676Z box:tasks setCompleted - 1874: {"result":null,"error":null,"percent":100}
2026-03-12T14:05:09.676Z box:tasks updating task 1874 with: {"completed":true,"result":null,"error":null,"percent":100}
2026-03-12T14:05:09.681Z Exiting with code 0
2026-03-12T14:05:09.681Z box:taskworker Task took 0.71 seconds
2026-03-12T14:05:09Z 2026-03-12 14:05:09,504 INFO Included extra file "/etc/supervisor/conf.d/redis-service.conf" during parsing
2026-03-12T14:05:09Z 2026-03-12 14:05:09,504 INFO Included extra file "/etc/supervisor/conf.d/redis.conf" during parsing
2026-03-12T14:05:09Z 2026-03-12 14:05:09,504 INFO Set uid to user 0 succeeded
2026-03-12T14:05:09Z 2026-03-12 14:05:09,514 CRIT Server 'inet_http_server' running without any HTTP authentication checking
2026-03-12T14:05:09Z 2026-03-12 14:05:09,514 INFO RPC interface 'supervisor' initialized
2026-03-12T14:05:09Z 2026-03-12 14:05:09,514 INFO supervisord started with pid 1
2026-03-12T14:05:09Z ==> Ensure directories
2026-03-12T14:05:09Z ==> Setting permissions
2026-03-12T14:05:09Z ==> Starting N8N
2026-03-12T14:05:09Z ==> Starting supervisor
2026-03-12T14:05:10.000Z => Healthcheck error: Error: connect ECONNREFUSED {redacted}:5678
2026-03-12T14:05:10Z 13:C 12 Mar 2026 14:05:10.542 # WARNING Memory overcommit must be enabled! Without it, a background save or replication may fail under low memory condition. Being disabled, it can also cause failures without low memory condition, see https://github.com/jemalloc/jemalloc/issues/1328. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.
2026-03-12T14:05:10Z 13:C 12 Mar 2026 14:05:10.542 * Configuration loaded
2026-03-12T14:05:10Z 13:C 12 Mar 2026 14:05:10.542 * Redis version=8.2.2, bits=64, commit=00000000, modified=1, pid=13, just started
2026-03-12T14:05:10Z 13:C 12 Mar 2026 14:05:10.542 * oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.543 * monotonic clock: POSIX clock_gettime
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.544 # Failed to write PID file: Permission denied
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.544 * Running mode=standalone, port=6379.
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * DB loaded from disk: 0.000 seconds
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * Done loading RDB, keys loaded: 0, keys expired: 0.
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * Loading RDB produced by version 8.2.2
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * RDB age 8 seconds
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * RDB memory usage when created 0.71 Mb
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * Ready to accept connections tcp
2026-03-12T14:05:10Z 13:M 12 Mar 2026 14:05:10.545 * Server initialized
2026-03-12T14:05:10Z 2026-03-12 14:05:10,518 INFO spawned: 'redis' with pid 13
2026-03-12T14:05:10Z 2026-03-12 14:05:10,522 INFO spawned: 'redis-service' with pid 14
2026-03-12T14:05:10Z Redis service endpoint listening on http://:::3000
2026-03-12T14:05:11Z 2026-03-12 14:05:11,740 INFO success: redis entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-03-12T14:05:11Z 2026-03-12 14:05:11,741 INFO success: redis-service entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-03-12T14:05:13Z Initializing n8n process
2026-03-12T14:05:13Z n8n Task Broker ready on 127.0.0.1, port 5679
2026-03-12T14:05:13Z n8n ready on ::, port 5678
2026-03-12T14:05:14Z Deprecation warning: The storage directory "/app/data/user/.n8n/binaryData" will be renamed to "/app/data/user/.n8n/storage" in n8n v3. To migrate now, set N8N_MIGRATE_FS_STORAGE_PATH=true. If you have a volume mounted at the old path, update your mount configuration after migration.
2026-03-12T14:05:14Z Failed to start Python task runner in internal mode. because its virtual environment is missing from this system. Launching a Python runner in internal mode is intended only for debugging and is not recommended for production. Users are encouraged to deploy in external mode. See: https://docs.n8n.io/hosting/configuration/task-runners/#setting-up-external-mode
2026-03-12T14:05:14Z [license SDK] Skipping renewal on init: license cert is not due for renewal
2026-03-12T14:05:15Z (Use `node --trace-deprecation ...` to show where the warning was created)
2026-03-12T14:05:15Z (node:1) [DEP0169] DeprecationWarning: `url.parse()` behavior is not standardized and prone to errors that have security implications. Use the WHATWG URL API instead. CVEs are not issued for `url.parse()` vulnerabilities.
2026-03-12T14:05:15Z Registered runner "JS Task Runner" (YbZvK9naPF-koKlaM82Kt)
026-03-12T14:05:18Z - My Workflow 1 (ID: bR91JeCWm5Mx4sFs)
etc etc
2026-03-12T14:05:18Z Building workflow dependency index...
2026-03-12T14:05:18Z Currently active workflows:
2026-03-12T14:05:18Z Found unfinished executions: 1767388, 1767389, 1767392, 1767414, 1767422
2026-03-12T14:05:18Z Start Active Workflows:
2026-03-12T14:05:18Z This could be due to a crash of an active workflow or a restart of n8n
2026-03-12T14:05:18Z Version: 2.11.3
2026-03-12T14:05:18Z [Recovery] Logs available, amended execution
2026-03-12T14:05:18Z [Recovery] Logs available, amended execution
2026-03-12T14:05:18Z [Recovery] Logs available, amended execution
2026-03-12T14:05:18Z [Recovery] Logs available, amended execution
2026-03-12T14:05:18Z [Recovery] Logs available, amended execution
2026-03-12T14:05:19Z (node:1) [DEP0040] DeprecationWarning: The `punycode` module is deprecated. Please use a userland alternative instead.
2026-03-12T14:05:19Z Activated workflow "My Workflow 1" (ID: 313JrHCjBD8mSLx8)
2026-03-12T14:05:19Z Issue on initial workflow activation try of "My Workflow 2" (ID: JaltokDBM25yzRQF) (startup)
2026-03-12T14:05:19Z libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory
etc etc
2026-03-12T14:05:21Z Activation of workflow "My Workflow" (TjFAytoNhfUA6QT9) did fail with error: "libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory" | retry in 2 seconds
2026-03-12T14:05:21Z Editor is now accessible via:
2026-03-12T14:05:21Z Press "o" to open in Browser.
2026-03-12T14:05:21Z Try to activate workflow "My Workflow" (TjFAytoNhfUA6QT9)
2026-03-12T14:05:21Z libc.musl-x86_64.so.1: cannot open shared object file: No such file or directory

[girish]

@umnz https://git.cloudron.io/packages/n8n-app/-/merge_requests/158