mod: first check then chown

[schenklklopfer]

As we restart our repomanager a lot, we noticed there is a chown running minutes.
Whilst this everything is very slow.

Checking is way faster than setting.

So first check, if it fails set the ownership new.

[lbr38]

Your proposal is not optimal, since a recursive chown is still applied if even a single file has the wrong permissions. What about:

find "$DIR" \( ! -uid "$UID" -o ! -gid "$GID" \) -exec chown "${UID}:${GID}" {} +

This will only modify files with incorrect permissions and avoid a recursive chown.

[schenklklopfer]

Yes, that's a way better thing for a few wrong files.
But it will take forever if all files needs to be fixed.

At the first instance I do not understand why this is needed at all.
Maybe there are cases, where all files needs fixing?
But only a few ones?

I do not know which case that could be...

Basically I just want to avoid the permanent chown on our half TB of files via slow NFSv3... This takes about 15 minutes.
And slows down the whole system after a restart of the container.

[lbr38]

But it will take forever if all files needs to be fixed.

If all files need to be fixed, so they have to be fixed no matter how much time it takes.. otherwise the app will not work. But that's in the worst-case scenario. In fact, permissions on files should never change. This chown is there as a safeguard, in case someone changes the file permissions by mistake. In this case, a restart of the container will fix the permissions automatically.

I understand this could be a performance issue.

Did you tried to launch the command I gave you in the previous message on your instance to see what time it takes to execute?

[schenklklopfer]

Not yet.
At the moment we test other performance improvements.

But next week I can test your command on our production.