Was the 9.10.1-U1 update tested? Update failure discussion

[JoshDW19]

This is going to cover a wide user base. Asking users to wait (especially when there is a critical Samba CVE) or asking people to remove any 'media' account they may have created, possibly breaking workflows, access, etc. in my view should not be an acceptable answer.

Would it be a better idea to pull the update and re-issue it when it's fixed?

I want to make sure we're on the same page here. If you manually remove the 'media' account then run the update it will then re-create the 'media' user for you automatically. This may have already been understood, but I wanted to clarify to make sure. It's not a perfect solution. Just a workaround for now if users would indeed like to update.

 

[m0nkey_]

I want to make sure we're on the same page here. If you manually remove the 'media' account then run the update it will then re-create the 'media' user for you automatically. This may have already been understood, but I wanted to clarify to make sure. It's not a perfect solution. Just a workaround for now if users would indeed like to update.

I understand what is being said. However, when the 'media' user is created, it's likely going to be configured using a different UID, with the possibility of breaking permissions in the process.

In future, when iX is adding a built-in user account, would be be an idea to prefix the name with ix_ or something?

 

[Ericloewe]

The second issue is related to UEFI installs only (And not all of them, I still cannot reproduce it here locally). It too has been fixed in source and will be in -U2. We are looking at -U2 being in the next ~30 days or so. Anyway, thanks for the bug reports, we appreciate your help in tracking these down.

To clarify: That second issue also causes a similar error upon upgrading, or is it something else?

 

[Ericloewe]

I understand what is being said. However, when the 'media' user is created, it's likely going to be configured using a different UID, with the possibility of breaking permissions in the process.

In future, when iX is adding a built-in user account, would be be an idea to prefix the name with ix_ or something?

It also seems like an odd user to add to the system, but I've been a bit confused with the planning for 9.10 - I'll have to investigate that when things calm down.

 

[anodos]

This is going to cover a wide user base. Asking users to wait (especially when there is a critical Samba CVE) or asking people to remove any 'media' account they may have created, possibly breaking workflows, access, etc. in my view should not be an acceptable answer.

Would it be a better idea to pull the update and re-issue it when it's fixed?

Critical Samba CVE?

 

[JoshDW19]

I understand what is being said. However, when the 'media' user is created, it's likely going to be configured using a different UID, with the possibility of breaking permissions in the process.

In future, when iX is adding a built-in user account, would be be an idea to prefix the name with ix_ or something?

I think what it comes down to is this feature was just not implemented well. If the developers wanted to add a user account there are a number of ways they can check to see if the user is already there that wouldn't mess with your current configurations. If it was an arbitrary user that only the system referenced then I think the developers would have used a prefixed ix-media directory, but in this case the developers wanted to keep things simple in case users are occasionally browsing / typing the name of the directory.

I can't make any promises but I have heard whispers about a hot-fix for this issue in the near future.

Apologies for the trouble everyone. I'll continue to update this thread as information becomes available.

 

[INCSlayer]

What uid and gid will the new built in media user and group have?

 

[JoeB]

Just updated and had the same issues... My VBox jail and some other jails all failed to start..

I've rolled back !

This "update" should be PULLED.

 

[Cellobita]

I'm really hoping that all this is because most development resources are now committed to the upcoming 10 release - sure, FreeNAS is "for enthusiasts", "not to be used on mission critical systems", etc., etc., but the last few releases have been hair-raising, to say the least...

 

[metalliqaz]

I have the exact same issue:

Code:

django.db.utils.IntegrityError: UNIQUE constraint failed: account_bsdgroups.bsdgrp_group

It was caused by the 'fix' to issue 16263. I already have a media group on my box.

I'm seriously surprised this was allowed.

 

[rogerh]

On my system U1 breaks some but not all replication tasks. I don't really want to go back to U1 because reverting to 9.10.1 made the two broken replication tasks recover. (Rebooting didn't so it is probably a real bug.) But if others have had this problem I don't mind doing some cautious experiments with my only backup!

 

[Ericloewe]

I hope the dev team makes an effort to communicate major changes in the future. If needed, I'm sure people on the forum can help with this, but some input from the dev team is going to be essential.

Current plans for 9.10, in particular, are confusing when gleaned exclusively from bug tickets - and it would be nice to understand what the plans are, because it almost looks like the idea is to effectively migrate the FreeNAS 8/9 GUI to the FreeNAS 10 infrastructure, including the new and improved middleware.

As for the matter at hand, I don't understand why there was a need to add a user named "media" this late in 9.10's lifecycle. It goes against best practices of avoiding (likely) existing names or similar, confusion-prone ones.

 

[scurrier]

I'm having replication problems with my plex jail after updating.

Second time I've had a major issue after updating FreeNAS. pfSense is also FreeBSD based and I have never had an issue after years and years of upgrades. I wish it was the other way around because my data's way more important to me than my router.

 

[Stux]

I found the built in docs in the webgui broken after the update.

For instance, 10.4 Windows (CIFS) Shares

/docs/freenas_sharing.html#windows-cifs-shares

/docs/_images/cifs2.png seems to be missing

and if I try to open the frame in another tab, I get a 404.

 

[gpsguy]

BTW, I generated new PDF and ePub versions of the docs.


Sent from my iPhone using Tapatalk

 

[JoshDW19]

@Stux have you made a bug ticket yet on the broken documentation? It'd be really helpful if you could.

 

[wblock]

I found the built in docs in the webgui broken after the update.

For instance, 10.4 Windows (CIFS) Shares

/docs/freenas_sharing.html#windows-cifs-shares

/docs/_images/cifs2.png seems to be missing

and if I try to open the frame in another tab, I get a 404.

This is weird, because that now is Windows (SMB) Shares and the image is called cifs2a.png. Checked on FreeNAS-9.10.1-U1 (ff51a5d). If you are running the same version, maybe just refresh that page in the browser?

 

[Stux]

This is weird, because that now is Windows (SMB) Shares and the image is called cifs2a.png. Checked on FreeNAS-9.10.1-U1 (ff51a5d). If you are running the same version, maybe just refresh that page in the browser?

Thanks for the tip. It wasn't quite as simple as reloading the page, as that just reloaded the GUI.

From the Develop menu in Safari I chose "Empty Caches", and all of a sudden the docs have been updated. Quite a difference!

Of course, my concern is that the Empty Caches menu is a developer option in Safari. Is there anything that should've been done as part of the upgrade to invalidate the cached pages? Perhaps the TTL or whatever it is that HTML uses is excessive?

 

[Ericloewe]

Thanks for the tip. It wasn't quite as simple as reloading the page, as that just reloaded the GUI.

From the Develop menu in Safari I chose "Empty Caches", and all of a sudden the docs have been updated. Quite a difference!

Of course, my concern is that the Empty Caches menu is a developer option. Is there anything that should've been done as part of the upgrade to invalidate the cached pages?

You can try CTRL+F5, or whatever the OS X equivalent is, which forces a proper reload on most browsers.

 

[Stux]

You can try CTRL+F5, or whatever the OS X equivalent is, which forces a proper reload on most browsers.

Yeah, I tried that. Its cmd-r btw, but it just reloads the web gui's outer frame back to the system information, leaving the docs frame with stale content. I tried an option-cmd-reload for a more powerful reload too.