[Fri Aug 29 19:55:58 2025] mineru-api[3353866]: segfault at 562d29d8d950 ip 0000562d29d8d950 sp 00007ffd4069c578 error 15

[loveRhythm1990]

🔎 Search before asking | 提交之前请先搜索

• I have searched the MinerU Readme and found no similar bug report.
• I have searched the MinerU Issues and found no similar bug report.
• I have searched the MinerU Discussions and found no similar bug report.

🤖 Consult the online AI assistant for assistance | 在线 AI 助手咨询

• I have consulted the online AI assistant but was unable to obtain a solution to the issue.

Description of the bug | 错误描述

We have encountered this issue several times, all occurring during normal PDF parsing operations. This is the output from the dmesg log. The problem is that the container does not restart but becomes stuck.

[Fri Aug 29 19:55:58 2025] mineru-api[3353866]: segfault at 562d29d8d950 ip 0000562d29d8d950 sp 00007ffd4069c578 error 15

How to reproduce the bug | 如何复现

normal pdf parse

Operating System Mode | 操作系统类型

Linux

Operating System Version| 操作系统版本

Linux 10-60-98-55-dev 5.15.0-113-generic #123-Ubuntu SMP Mon Jun 10 08:16:17 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

Python version | Python 版本

3.10

Software version | 软件版本 (mineru --version)

最新版本：sglang v2.1.10

Device mode | 设备模式

cuda

[dosubot]

Hi @loveRhythm1990! I'm Dosu and I’m helping the MinerU team.

This segfault and container hang in MinerU 2.0.x with CUDA is likely due to a known GPU memory leak and resource management issue during PDF parsing, especially in the MFR Predict/model inference step. This causes GPU memory usage to grow until exhaustion, leading to process instability or a stuck container. The issue is fixed in MinerU 2.1.2 and later—upgrading is strongly recommended for stability and better resource handling (source).

If upgrading is not immediately possible, you can try these mitigations: reduce GPU memory usage with the --mem-fraction-static parameter in your Docker Compose config, limit the number of concurrent PDF parses, or switch to CPU mode for testing (source). Also, batch PDF parsing can be interrupted by corrupted files, so try smaller batches and isolate problematic PDFs if you see frequent failures (source).

There are no direct fixes for segfaults or container hangs in recent PRs, but many stability improvements have been made in newer versions. If you need to debug further, collecting core dumps and checking for problematic PDFs or concurrency patterns may help.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[myhloli]

检查下内存或者显存有没有满，gpu有没有满负载工作