Abnormal disk IO in version 1.13.1 #450

New Issue

MrUnknownDE · 2026-04-05T17:10:54+02:00

MrUnknownDE commented

2026-04-05 17:10:54 +02:00

Originally created by @asardaes on 12/21/2025

Describe the Bug

I'm not sure if this is related to #2120, so I figured I would report it to be sure. My Pangolin container hasn't suffered from OOM kills, but it has been lagging significantly; I set the memory limit of the container to 320M. Now I looked at the disk metrics from my VPS and I see something like this:

Environment

OS Type & Version: Ubuntu 24.04.3 LTS
Pangolin Version: 1.13.1 with geolite2 DB
Gerbil Version: 1.3.0
Traefik Version: 3.6.5
Newt Version: 1.7.0
Olm Version: 1.2.0

Crowdsec is not running on the VM hosting Pangolin.

To Reproduce

You'll notice a couple spikes in the screenshot I posted. In this experiment I did the following:

Enable site A which only has a private resource configured. In this machine Newt is running directly on the VM through systemd, no container. The disk read spike after this appears to have normalized.
Enable machine X. This is a Docker container using Olm to connect, and it's the machine that gets access to the private resource from site A. Very small spike seen in the screenshot.
Enable site B with multiple public resources. This is another Docker container running Newt. This caused the read throughput to skyrocket and stay there until I disabled the site again.

I wonder if the memory constraints from Pangolin's container make it flush caches continuously and reload them from disk.

Expected Behavior

Sane disk IO.

*Originally created by @asardaes on 12/21/2025* ### Describe the Bug I'm not sure if this is related to #2120, so I figured I would report it to be sure. My Pangolin container hasn't suffered from OOM kills, but it has been lagging significantly; I set the memory limit of the container to 320M. Now I looked at the disk metrics from my VPS and I see something like this: <img width="377" height="354" alt="Image" src="https://github.com/user-attachments/assets/80d0e806-c47c-42f1-b7d8-9a4003630250" /> ### Environment - OS Type & Version: Ubuntu 24.04.3 LTS - Pangolin Version: 1.13.1 with geolite2 DB - Gerbil Version: 1.3.0 - Traefik Version: 3.6.5 - Newt Version: 1.7.0 - Olm Version: 1.2.0 Crowdsec is *not* running on the VM hosting Pangolin. ### To Reproduce You'll notice a couple spikes in the screenshot I posted. In this experiment I did the following: 1. Enable site A which only has a private resource configured. In this machine Newt is running directly on the VM through systemd, no container. The disk read spike after this appears to have normalized. 2. Enable machine X. This is a Docker container using Olm to connect, and it's the machine that gets access to the private resource from site A. Very small spike seen in the screenshot. 3. Enable site B with multiple public resources. This is another Docker container running Newt. This caused the read throughput to skyrocket and stay there until I disabled the site again. I wonder if the memory constraints from Pangolin's container make it flush caches continuously and reload them from disk. ### Expected Behavior Sane disk IO.

MrUnknownDE closed this issue

2026-04-05 17:10:54 +02:00

Sign in to join this conversation.

Branches Tags

main

dev

dependabot/npm_and_yarn/dev-minor-updates-b4e5d6b9c5

revert-2766-feature/systemd-install-instructions

dependabot/npm_and_yarn/prod-patch-updates-05702d39f2

dependabot/npm_and_yarn/next-16.2.1

dependabot/npm_and_yarn/recharts-3.8.1

alerting-rules

private-site-ha

dependabot/docker/docker/library/node-25-slim

ssh

delete-account

msg-delivery

org-only-idp

cicd

patch

site-targets-auto-login

No Label

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github/pangolin#450