Skip to content
Problems.

Jobs and services run out of memory

DevOps and infra7 posts from 7 people6d active+11 post in the last 7 days, 1 the 7 days before (steady)

Posts per day

Posts per day7 posts, Aug 28 to Sep 25

The posts behind it

7, newest first
PostDate
Consider a dedicated allocator for Windows embedded DuckDB workloads### What happens? ## Background I recently investigated a Windows performance problem in DuckDB when it is embedded in an R process hosted by Positron/RStudio. The investigation started from a substantial loss of DuckDB parallel performance after repeated workloads, and the...duckdb/duckdbkoenniemSep 214 days ago
[platform] "Exit code file empty" fails an attempt whose destination logged "Completed integration"## What The orchestrator reads the destination's exit-code file immediately after the destination logs `Completed integration`. When the file is not written yet, the orchestrator throws `IllegalStateException: Exit code file empty` and fails the attempt, despite having already...airbytehq/airbytejtrutySep 169 days ago
Multi-main: workflows lose their schedule and poll triggers when leadership is taken over mid-activation...re-saving each affected workflow was the only way to recover it. Two things make this hard to notice: 1. There is no operator-visible signal. The workflow is active everywhere you would look, and the activation-errors endpoint is empty. 2. The one log line that records it is at...n8n-io/n8nfluty84Sep 72 weeks ago
Spark: Async micro-batch preload does not stop when either limit is reached# Spark: Async micro-batch preload does not stop when either limit is reached ### Apache Iceberg version `main` (`0107f19a4a`); present in Spark 3.5, 4.0, and 4.1. ### Query engine Spark Structured Streaming with `async-micro-batch-planning-enabled=true`. ### Description...apache/icebergiamrajatranaSep 43 weeks ago
Live DataFrames can change silently after `collect()` under memory pressure## Summary Under sustained direct reclaim in a memory cgroup, a DataFrame matched its reference immediately after `scan_parquet().collect()`, then the same still-live DataFrame changed seconds later. No exception or warning was raised. Changes include all-zero 4 KiB regions and...pola-rs/polarstaozuoqiaoAug 313 weeks ago
Streaming engine: `.over()` segfaults and returns incorrect results under cgroup memory pressure...allows projection pushdown to remove the window expressions before reaching the failing path. - 12 million rows. Tests with 4 million and 6 million rows produced no failures in 12 runs under verified reclaim. No specific thread count is required, but it affects memory demand and...pola-rs/polarschuqCTCAug 284 weeks ago
Turbopack dev: PostCSS worker runs with the wrong cwd when an ancestor package.json exists, then spawns workers past the pool bound until OOM...bound. ### Expected The PostCSS worker resolves from the project directory, and a failing worker does not cause unbounded process growth. ### Isolating variables Same machine, same project, same page (byte-identical output): | | processes | memory | result | |---|---|---|---| |...vercel/next.jsPaliy0Aug 284 weeks ago

Companies and products named

Company or productPosts naming it
Polars2
BigQuery1
Airbyte1
DuckDB1
Positron1
Segment1
About this problem

Evidence

7 posts from 7 people in 6 places, about 2 a week over 25 days. Mostly on pola-rs/polars, airbytehq/airbyte, duckdb/duckdb. Tools named alongside: Polars, BigQuery, Airbyte, DuckDB.

Frustration Frustration 0 of 3· Seen on GitHub issues

How it was grouped

Posts that state a pain and match the "oom" rule. First post Aug 28, 2026, latest Sep 21, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

The same problem elsewhere

History

  • 2026-09-25 Added: Jobs and services run out of memory (7 posts)

Rising problems by email

Mondays: the problems in data, tech and AI that grew fastest that week.

Double opt-in. Unsubscribe any time.