Skip to content

Too slow docker boot on NFS mounts #6554

@druppy

Description

@druppy

Describe the bug

I have been using the docket image as part of a new HELM v3 upgrade I made (I have no idea where to publish this as of now), and it works really good, except for a small detail.

When starting my docker image, it took a really long time to startup, and in fact the k8s cluster I used ended up terminating the container as the live check only works when the service are running.

I took a deep look inside the container, and found that there where a chown -R $USER:$USER /data command running. This would normally not be a problem for small and physical mounts, but my setup was a really large (20k+ files) NFS mount.

I have therefor made a pull request #6553 , that change the start.sh to test if /data have the ownership we expect, and then only run the chown command when needed. There could be an argument about what happens if one of the files down the /data tree have changed ownership, and yes ... that would be a problem. But I assume that the chown command are here to ease migration, and not to keep file ownership in sync.

I really hope this could get pulled, to allow for better NFS integration.

Gogs version and commit
master

Git version
2.30.2

Operating system
Debian Linux

Database
PostgrSQL

To Reproduce

Steps to reproduce the behavior:

  1. setup docker image using a large NFS mount as /data
  2. start the container
  3. observe really long boot (depending on the location on your NFS server.

Can you reproduce the bug at https://try.gogs.io?
Nope, this is boot relevant

Expected behavior
Fast boot

Actual behavior
Really long startup time

Metadata

Metadata

Assignees

No one assigned

    Labels

    ⚓️ dockerCategorizes as related to Docker💊 bugSomething isn't working📦 deploymentRelated to deployments

    Type

    No type

    Projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions