Data and AI struggles

2,114 problems · 377 companies

Outages and change

SQLite upgrades break working setups

20 posts from 19 people15d activeSteady, +1 this week

Posts per day

Posts per day20 posts, Sep 1 to Nov 21

The posts behind it

GitHub issuesPrefectHQ/prefect · bbaJay749Nov 17

SQLite lock errors in worker channel heartbeat persistence

### Bug summary With a SQLite database, the Prefect server sometimes logs this error while workers are connected through the worker channel. In the traceback below, paths, the work pool name, the worker name and UUIDs are replaced. The log timestamp is converted to UTC. **What I...

GitHub issuesn8n-io/n8n · tomasr78Nov 10

ConvertAPI credential test passes with an invalid API token

### Bug Description The ConvertAPI credential test says "Connection successful!" for any API token, even a wrong or empty one. Users find out only when a workflow runs and ConvertAPI answers with 401. The test sends `GET https://v2.convertapi.com/convert/docx/to/pdf` with...

1 replies
GitHub issuesfosrl/pangolin · fdesouza-stackNov 5

1.26.0 image: dashboard and integration API fail to start (Cannot find module 'next/dist/compiled/webpack/webpack-lib')

## Describe the bug After upgrading a self-hosted Community Edition install from 1.24.0 to 1.26.0, the database migrations complete successfully, but the Dashboard Web UI (port 3002) and the Integration API (port 3003) never start. The dashboard returns 502 through Traefik. The...

1 replies
GitHub issuesn8n-io/n8n · jeonghun917Nov 4

RabbitMQ node: queue bindings aren't awaited before the channel is returned

### Bug Description This happens at line 80 in packages/nodes-base/nodes/RabbitMQ/GenericFunctions.ts. forEach in options.binding.bindings.forEach(async (binding) => { ... bindQueue ... }) doesn't wait for the async callbacks. As a result, it returns the channel before bindQueue...

3 replies
GitHub issuesn8n-io/n8n · marcoo93Nov 3

Regular mode with a production concurrency limit: static data is still read before the concurrency wait (follow-up to #39653)

### Bug Description #39668 (released in 2.43.0) fixed the queue-mode case reported in #39653 by passing `loadStaticData` to `enqueueExecution()`. The same change in #37701 also affects regular (main) mode when `N8N_CONCURRENCY_PRODUCTION_LIMIT` is set, and that path is unchanged...

3 replies
GitHub issuesrun-llama/llama_index · bossjoker1Nov 1

Memory.reset() during in-flight finalization can resurrect old messages

### Summary `Memory.reset()` can return an empty history while an `aput_messages()` or agent finalization that started before the reset is still in flight; the older write then lands after the reset and the old context is visible again. I reproduced this on `llama-index-core`...

GitHub issuesn8n-io/n8n · mayur-dhavanOct 25

isBetween() with date-only string bounds drops the written day in non-UTC workflow timezones

### Bug Description In a workflow expression, `DateTime.local(2023, 6, 16, 23, 0).isBetween('2023-06-16', '2023-06-17')` returns `false` when the workflow timezone is west of UTC (for example America/New_York), even though 23:00 on June 16 is inside the written range. Date-only...

1 replies
GitHub issuesn8n-io/n8n · indexGinOct 25

String .extractUrl() includes trailing sentence punctuation in the extracted URL

### Bug Description `.extractUrl()` includes punctuation that belongs to the surrounding sentence, not the URL itself. When a URL appears at the end of a sentence, before a comma, or inside parentheses, the extracted value keeps the trailing marks: - `"Docs:...

2 replies
GitHub issuesn8n-io/n8n · rmyndharisOct 25

Static data written in a webhook trigger's webhook() is not persisted since 2.19.2

## Bug Description Static data written from a webhook trigger's `webhook()` is not persisted on n8n 2.19.2 and later. The write is visible during the execution, but the next delivery starts from the previously stored state, so counters, cursors and de-duplication records kept in...

4 replies
GitHub issuesn8n-io/n8n · remkostasOct 25

Schedule Trigger runs daily workflows twice on DST fall-back night (Europe/Berlin)

...this for America/New_York with daily jobs at 02:xx. US clocks go back on 1 Nov 2026. Workaround: I moved the jobs where a second run hurts to 04:xx. ### To Reproduce 1. Self-hosted n8n with GENERIC_TIMEZONE=Europe/Berlin and TZ=Europe/Berlin, N8N_SCHEDULER_ENABLED not set. 2....

2 replies
GitHub issuesn8n-io/n8n · JohanJiantaOct 24

SFTP: "Timed out while waiting for handshake" after upgrading to 2.44.6 (Node 26 slow 8192-bit DH group exchange)

...and preferred bits capped at 4096 | ~6.7 s | Debug lines on Node 26 (ms since start): **Workaround:** patching `ssh2/lib/protocol/kex.js` in a custom Dockerfile to cap `GEX_MAX_BITS` and the preferred size from `dhEstimate()` at 4096. The credential connects again: **Suggested...

2 replies
GitHub issuesn8n-io/n8n · DSAntonio08Oct 24

Compression Node: gzip file names get mangled ("archive.gzip" becomes "archiveip") or cut at the first dot

### Bug Description The Compression node handles gzip file names with plain string operations that break in two ways. **1. Custom file name ending in `.gzip`** (Compress, gzip, node version > 1) The first `replace` already removes `.gz` from `.gzip`, so `"archive.gzip"` becomes...

1 replies
GitHub issuesrun-llama/llama_index · emme1tOct 23

[Bug]: SimplePropertyGraphStore.from_dict clears the caller's node snapshot

### Bug description `SimplePropertyGraphStore.from_dict(data)` replaces `data["nodes"]` with `{}` while preparing manual node deserialization. This modifies the caller's snapshot in place, even if graph validation later raises an exception. Loading or serializing the same...

GitHub issuesn8n-io/n8n · mihai-mmOct 21

Instance AI / Agents: thread keeps using a deleted sandbox after the runner restarts ("sandbox not found") and never recreates it

...for `409 sandbox_restarted`. Workspace files are lost, but the thread keeps working. Workaround we use for now: a cron job on the host that records the ids from `GET /sandboxes` every minute and re-POSTs them when the runner's start time changes. ### Debug Info Not copied from...

2 replies
GitHub issuesPrefectHQ/prefect · web3wannabeOct 20

SQLite lock errors while updating automation buckets during event processing

## Bug summary A self-hosted Prefect server using SQLite can fail while processing an event that updates an `automation_bucket` row: This has been observed under concurrent database writes with SQLite in WAL mode. The failure occurs in the event-processing path while evaluating...

1 replies
GitHub issuescrewAIInc/crewAI · lorenzejayOct 13

[FEATURE] Add exact historical Flow state lookup scoped to its UUID

## Python PR scope Add a read-only method to retrieve one historical state row by its stable row ID and Flow UUID, alongside the history listing API. **Difficulty:** easy. **Allocation index:** #8118. **Project:** #8121. ## Cohort allocation This ticket is reserved for...

2 replies
GitHub issuescrewAIInc/crewAI · lorenzejayOct 13

[Project] Inspect historical SQLite Flow persistence states

...Reuse existing interfaces, fixtures, and dependencies. Runtime behavior changes need a failing-before/passing-after regression. Do not compete with already-open fixes; note dependencies and coordinate shared files. Required checks run without paid model calls. ## Starting points...

GitHub issuesn8n-io/n8n · CentigradePatrickAndreOct 12

If-node in sub-workflow causes properties of previous node to become not found when all items go the false path

...items in an if-node use the false path and a merge node is behind it, the merge node is unable to execute. ### To Reproduce [main-workflow.json](https://github.com/user-attachments/files/33341808/main-workflow.json)...

3 replies
GitHub issuesrun-llama/llama_index · Harsh23KashyapSep 29

SQLDatabase schema prefixing changes FROM/JOIN text inside SQL string literals

### What happens When `SQLDatabase` is configured with a schema, `run_sql()` rewrites `FROM` and `JOIN` matches throughout the query string, including inside quoted SQL literals. This changes returned data, even though the actual table reference is prefixed correctly. SQL...

1 replies
GitHub issuesvllm-project/vllm · akalin9507Sep 1

[RFC] Bounded, restart-safe capacity management for the filesystem KV offload tier

### Motivation. The current filesystem KV offload tier (`type: "fs"`, `TieringOffloadingSpec`) has no configurable disk capacity limit. Persisted KV blocks accumulate indefinitely and may exhaust the backing disk in long-running deployments. The existing configuration also...

2 reactions

Companies and products named

20 posts
11 posts
7 posts
5 posts
3 posts
2 posts
About this problem

Evidence

20 posts from 19 people in 6 places, about 2 a week over 78 days. Mostly on n8n-io/n8n, run-llama/llama_index, crewAIInc/crewAI. Tools named alongside: n8n, Docker, GitHub, LlamaIndex.

Frustration Frustration 0 of 3· Seen on GitHub issues

Summary, model-written

SQLite users face data corruption from query rewriters modifying quoted literals, unconstrained disk accumulation in storage tiers, and workflow node execution failures.

How it was grouped

Posts that state a pain and match the "breaking-changes" rule about SQLite. First post Sep 1, 2026, latest Nov 17, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

The same problem elsewhere

GitHub upgrades break working setups

109 posts from 94 people on GitHub issues, Hacker News, Stack Overflow.

Upgrades and new releases break working setups

94 posts from 76 people on GitHub issues, Reddit, App Store reviews.

Claude Code upgrades break working setups

82 posts from 73 people on GitHub issues, Hacker News, Reddit.

Codex upgrades break working setups

72 posts from 65 people on GitHub issues, Reddit.

DuckDB upgrades break working setups

55 posts from 42 people on GitHub issues.

Claude upgrades break working setups

46 posts from 44 people on GitHub issues, Reddit, Hacker News.

Ollama upgrades break working setups

40 posts from 20 people on GitHub issues.

ClickHouse upgrades break working setups

37 posts from 21 people on GitHub issues.

OpenAI upgrades break working setups

32 posts from 32 people on GitHub issues, Reddit, Community forums.

Anthropic upgrades break working setups

30 posts from 24 people on GitHub issues.

ChatGPT upgrades break working setups

29 posts from 29 people on GitHub issues, Community forums, App Store reviews.

PyTorch upgrades break working setups

29 posts from 23 people on GitHub issues, Reddit.

Other SQLite problems

SQLite users keep filing bugs

Top post: "Form Trigger execution succeeds but editor remains stuck on Waiting for you to submit the form", one of 42 posts from 39 people on GitHub issues, Reddit.

Trouble with SQLite node

Top post: "HTTP Request node (v3+) re-decodes valid UTF-8 JSON responses as GB18030 when the text contains Latin-1 range", one of 6 posts from 6 people on GitHub issues.

SQLite rate limits and usage caps block work

3 posts from 3 people on GitHub issues, Mastodon.

Trouble with SQLite returns

Top post: "A Jinja expression that cannot be evaluated returns 500 instead of a template error", one of 3 posts from 3 people on GitHub issues.

Trouble with SQLite response

Top post: "Provide More Context When the Error Something Went Wrong Before I Could Finish That Response Occurs", one of 2 posts from 2 people on GitHub issues. Unverified.

SQLite login, SSO or OAuth breaks

2 posts from 2 people on GitHub issues. Unverified.

SQLite lock-in makes switching costly

2 posts from 2 people on GitHub issues. Unverified.

Trouble with SQLite contains

Top post: "ReadWriteFile and WriteBinaryFile tests fail on macOS because the temp path contains a symlink", one of 2 posts from 2 people on GitHub issues. Unverified.

Experiments on SQLite are hard to reproduce and track

2 posts from 2 people on GitHub issues. Unverified.

Trouble with SQLite mode

One post on GitHub issues: "Use WAL mode for the REST fixture's SQLite backend". Unverified.

Trouble with SQLite http

One post on GitHub issues: "HTTP Request parses top-level JSON strings twice in Autodetect response mode". Unverified.

History

Momentum went from fading to steady.

Nov 20, 2026

Momentum went from steady to fading.

Nov 11, 2026

Momentum went from rising to steady.

Nov 10, 2026

Momentum went from steady to rising.

Nov 9, 2026

Momentum went from fading to steady.

Nov 7, 2026

Momentum went from steady to fading.

Nov 1, 2026

Momentum went from rising to steady.

Oct 31, 2026

Summary written by a model: SQLite users face data corruption from query rewriters modifying quoted literals, unconstrained disk accumulation in storage tiers, and workflow node execution failures.

Oct 28, 2026

Momentum went from steady to rising.

Oct 27, 2026

Momentum went from fading to steady.

Oct 26, 2026

Momentum went from steady to fading.

Oct 20, 2026

Corroborated: 3 or more posts from 2 or more people or places.

Oct 17, 2026