Instance unhealthy after v0.14.0: MeiliSearchApiError: The provided API key is invalid.

[mmanjos]

I seem to be having some issues after upgrading to this morning's v0.14.0 release

I'm using the latest docker-compose.yml from main so the migration to meilisearch v1.11.3 was included:

(...)
    container_name: wanderer-search
    image: getmeili/meilisearch:v1.11.3
(...)

Basically, my docker-compose.yml is the same as main but with these modifications:

• wanderer-web environment variable ORIGIN is set to the hostname
• PUBLIC_DISABLE_SIGNUP set to true
• MEILI_MASTER_KEY set to a persistent, secret value using cenv

here's how that looks:

x-common-env: &cenv                                                                                     
  MEILI_URL: http://search:7700
  MEILI_MASTER_KEY: secretvalue

After starting the new docker-compose stack, it looks like meilisearch starts up cleanly:

After trying to load /trails I get this error in the logs:

wanderer-web     | MeiliSearchApiError: The provided API key is invalid.
wanderer-web     |     at /app/node_modules/meilisearch/dist/bundles/meilisearch.cjs.js:378:19
wanderer-web     |     at Generator.next (<anonymous>)
wanderer-web     |     at fulfilled (/app/node_modules/meilisearch/dist/bundles/meilisearch.cjs.js:298:58)
wanderer-web     |     at process.processTicksAndRejections (node:internal/process/task_queues:105:5) {
wanderer-web     |   code: 'invalid_api_key',
wanderer-web     |   type: 'auth',
wanderer-web     |   link: 'https://docs.meilisearch.com/errors#invalid_api_key',
wanderer-web     |   httpStatus: 403
wanderer-web     | }

[mmanjos]

(and this secretvalue did not change between versions - I keep that persistent in my environment)

[mmanjos]

hmm, it seems like I get that error if I try to access wanderer through my nginx reverse proxy (SSL)

if I go directly to the web container (tcp/3000) I get this error in the logs instead:

wanderer-web     | MeiliSearchApiError: Attribute `distance` is not filterable. This index does not have configured filterable attributes.
wanderer-web     | 1:9 distance >= 0 AND elevation_gain >= 0 AND elevation_loss >= 0 AND difficulty IN [easy,moderate,difficult] AND ((public = true OR author = undefined) OR shares = undefined)
wanderer-web     |     at /app/node_modules/meilisearch/dist/bundles/meilisearch.cjs.js:378:19
wanderer-web     |     at Generator.next (<anonymous>)
wanderer-web     |     at fulfilled (/app/node_modules/meilisearch/dist/bundles/meilisearch.cjs.js:298:58)
wanderer-web     |     at process.processTicksAndRejections (node:internal/process/task_queues:105:5) {
wanderer-web     |   code: 'invalid_search_filter',
wanderer-web     |   type: 'invalid_request',
wanderer-web     |   link: 'https://docs.meilisearch.com/errors#invalid_search_filter',
wanderer-web     |   httpStatus: 400
wanderer-web     | }

[Flomp]

This is likely due to a change that makes lists searchable. This required me to update the meilisearch tokens for every user. This migration should have made sure that all tokens are automatically updated.
Ah, but maybe you need to logout and login to invalidate your token. Could you try that please?

[mmanjos]

Very strange - I tried logging out and logging back in, but it didn't solve the problem.

I tried restoring a backup of all the volumes from before I upgraded, then started the stack again and this time it seems like the migration(s) worked fine and the instance is healthy again. Very weird - if I find anything in the logs about the first upgrade I'll post here, but for now it seems like it's working again ¯_(ツ)_/¯