Deployment kludges?

[kbaegis]

The deployment process seems unstable when using an external mongodb instance, or a private S3 compatible endpoint.

Attempted to deploy per: https://github.com/hcengineering/platform/blob/be6bd0d944ec60110de3b71644132c0e5830c0ba/server/server-storage/src/starter.ts#L21

And attempted to use a standard docker.io/mongo deployment, and subsequently mongo:7-jammy with identical values.

All cases were without success and errors either during workspace startup or with basic connectivity.

Additionally tested out the postgres connection. Seems to have a datatype issue with UUIDs.

Are portions of your implementation tightly coupled to docker name resolution?

[kbaegis]

Postgres:
{
message: 'failed to process sql',
error: 'invalid input syntax for type uuid: "w-test-test4-67ffa3b0-ba1551903d-08fa64"',
cause: PostgresError: invalid input syntax for type uuid: "w-test-test4-67ffa3b0-ba1551903d-08fa64"
at UD (/usr/src/app/bundle.js:707:270)
at un (/usr/src/app/bundle.js:697:1431)
at Socket.la (/usr/src/app/bundle.js:696:309982)
at Socket.emit (node:events:524:28)
at addChunk (node:internal/streams/readable:561:12)
at readableAddChunkPushByteMode (node:internal/streams/readable:512:3)
at Readable.push (node:internal/streams/readable:392:5)
at TCP.onStreamRead (node:internal/stream_base_commons:189:23) {
severity_local: 'ERROR',
severity: 'ERROR',
code: '22P02',
where: "unnamed portal parameter $1 = '...'",
file: 'uuid.c',
line: '133',
routine: 'string_to_uuid'
}
}

[kbaegis]

Using external MongoDB: {"err":{"message":"Unexpected token 'I', "Internal S"... is not valid JSON","stack":"SyntaxError: Unexpected token 'I', "Internal S"... is not valid JSON\n at JSON.parse ()\n at parseJSONFromBytes (node:internal/deps/undici/undici:5731:19)\n at successSteps (node:internal/deps/undici/undici:5712:27)\n at fullyReadBody (node:internal/deps/undici/undici:4609:9)\n at process.processTicksAndRejections (node:internal/process/task_queues:105:5)\n at async consumeBody (node:internal/deps/undici/undici:5721:7)\n at async getPendingWorkspace (/usr/src/app/bundle.js:299:1210175)\n at async /usr/src/app/bundle.js:788:306213\n at async ctr.start (/usr/src/app/bundle.js:788:306152)"},"level":"error","message":"Error getting pending workspace:","timestamp":"2025-04-16T12:47:59.036Z"}

[kbaegis]

  - DB_URL=mongodb://192.168.5.1:27017
  - MONGO_URL=mongodb://192.168.5.1:27017
  - STORAGE_CONFIG=s3|https://s3.local:9999?accessKey=<removed>&secretKey=<removed>&region=none

[kbaegis]

Got Mongo working by using a new bridged IP; postgres and s3 functionality still failing. https://hulycommunity.slack.com/archives/C06DVEUCCLW/p1744832243979889?thread_ts=1744829585.620939&cid=C06DVEUCCLW

TLDR: error logs from frontend indicate that despite S3 being configured for port 9999, port 443 is still in use.

[kbaegis]

I was able to isolate this exclusively to the domain of hcengineering's codebase. I can't deploy this against a proxy for S3 with a non tcp/443 port, nor for a S3 compatible service on a distinct port, and this is due to the hcengineering's code.

I would greatly appreciate the flexibility to implement against s3 compatible endpoints.

[kbaegis]

I already had an nginx frontend available to remap through, so I ended up leveraging extra_hosts: { <host>: <ip> } mapping to get around this.

Now the DNS based buckets are the only blocker for me.

An error was encountered in a non-retryable streaming request.
{"code":"ENOTFOUND","errno":-3008,"hostname":"w-test-test-67ffd3ad-84606d2767-614daf.s3.local","level":"error","message":"error-post-files","syscall":"getaddrinfo","timestamp":"2025-04-17T00:52:10.939Z"}
{"level":"info","message":"::ffff:172.29.0.1 - POST /files HTTP/1.0 500 - - 3.629 ms\n","name":"requests","timestamp":"2025-04-17T00:52:10.939Z"}

Would be awesome if we could remap this from .s3.url to s3.url/