systemd mounts
 

Hi there,

A few weeks ago I came here with a problem that not all file systems being mounted at boot time (https://www.linuxquestions.org/quest...ot-4175677301/)

Turns out the probable cause, as suggested by @The Squash and @syg00 was the order the system mounts file systems and the fs that I wanted to mount depends on other fs to be mounted first (/home). This cause an erratic behavior. Sometimes it get mounted other times it don't, depending on which get mounted first.

So, to fix that I moved my mounts from /etc/fstab to a unit file to get mounted by systemd.

And created several unit files like this one. The others are identical, except by the fs itself.
The problem is it works only 1 or 2 in 7 boots. Very annoying. When it don't work after the boot, "systemctl start home-miguel-Documents.mount" mounts the fs as expected, no warnings, no messages....

I have no indication of failure, other than that strange pattern Mounted/Umounting/mount: succeeded/Umount at end of status and, of course, reported by "journalctl -b -u home-miguel-Documents.mount"...

I already tried with and without "After=home.mount" , no effect.
The current mount targets in my system (openSuSE tumbleweed) are:
(this one: Documents in status active is because I issued the start command in a console. Before it was inactive, like the others)
What I am doing wrong ?
Why I can not get theses extra filesystems get mounted at boot time 100% of time ?

I appreciate any comments,

cheeers,

Why not put the mount commands in your bashrc file ( or whatever startup files u are using ) so that the disks are mounted as soon as you login

Probably not helpful for your problem... I have a system running Leap (using nothing but ext4) and another running Tumbleweed (ext4 on md for "/", btrfs for everything else) and neither have had any problems mounting everything that's listed in /etc/fstab. Everything using ext4 is mounted by label. Everything btrfs is mounted by device name. Both up-to-date as of last weekend but both systems have mounted everything flawlessly since installation.

Note: None of my btrfs subvolumes use that "@" notation in /etc/fstab. Your's all do. I'm wondering if that's an important difference. I'm not suggesting you change that; just noting that it's different. (I've yet to find anything in the manpages that discusses its significance; only a couple of web pages that claim it's "part of the mount point name" without actually explaining how or why.)

Cheers...

This is not a fix, it is a workaround.
I do not want a workaround, I want to understand why the things are not working as it was supposed to work, and them, fix it.

As I mentioned at my post, my first approach was listing the FS in fstab, but, due to hierarchical dependencies, it not worked either.

post edit: and about the '@' that is from the installer. I left to the installer create the default disk setup; Looks like is some SuSE convention, and as far I could understand it is just a ordinary character, with no special meaning to btrf, but it was created 'as it' by the tumbleweed installer.

There are "Before" and "After" unit directives, that may work:
https://fedoramagazine.org/systemd-u...ies-and-order/

Full disclosure, I never tried this to mount filesystems - only to start service which was on different FS.

Hi

Systemd will mount with forking - kind of like "mount -a -F" vs. "mount -a".

But you can set up dependencies in fstab - adding x-systemd.requires-mounts-for= for each mount point that depends on another mount point. Here is some info:

https://www.golinuxcloud.com/mount-f...order-systemd/

That could be part of the problem you're seeing. My all-ext4 system has filesystems mounted on filesystems (sort of grew organically; I likely wouldn't do it that way again) but mounts haven't been an issue.

The system using btrfs is simpler but looks a lot like what you have. The major difference I see is your subvolume naming:
I kept subvolume names simple (short, no special characters, less typing).

Yes. That's the explanation I've seen---that it was an installer default/convention. Special characters like that make me closely read the manpages looking for some special meaning. I added the btrfs storage after the initial install so did everything by hand. None of the documentation I read mentioned "@" characters. The last time I saw "@" used in fstab was for NFS mounts ("/exported-fs@somenfshost" ... long time ago)

Yes, I am already tried "After=home.mount", no effect.

Traditionally fstab mounts are done in sequence, top-down.
If systemd doesn't do that then there is a compatibility bug in systemd.

All my testing indicates systemd handles the ordering just fine. I was mis-attributed above - my feeling as expressed in the referenced thread is that it's a mechanical issue with the spinning disk

^ you mean one would need to add a delay to give the mounted disk time to actually become accessible?

Indeed as per the other thread, and also by @Guttorm above.

looks strange for me, mount itself should check the filesystem itself somehow (existence, type, dirty flag). Inaccessible filesystem cannot be mounted.

Hi, yes, you're right. Sorry.

About your specific comment, I think that it was the case, than I will not have the problem, because the fast home fs will be mounted first and the slow disk is mounted after a while, after the home is already ready to receive the sub-mounts, isn't ?

PS: Do you have hierarchical mounts, like me ? /home on a disk and /home/your-user/whatever in another disk ?