AFP Time Machine Sharing

[MtK]

Hey,
I have an AFP Sharing with TimeMachine enabled, which is accessible from the OSx's on the network.
TimeMachine sees the FN Share and tries to use it but fails.

after that, none of the AFP shares are available/reachable...


the log during this session is:

Code:

Feb 24 13:35:10 FN afpd[8851]: Login by mtk (AFP3.4)
Feb 24 13:35:11 FN afpd[8851]: AFP logout by mtk
Feb 24 13:35:11 FN afpd[8851]: AFP statistics: 1.55 KB read, 1.32 KB written
Feb 24 13:35:11 FN afpd[8851]: done

(not really showing a problem)

 

[Glorious1]

I don't know what the problem is, but if it were me, I would proceed as follows:

1. Turn the AFP service in FreeNAS off and back on, see if that resolves it. If not,
   
2. Remove the Time Machine destination from Time Machine preferences, delete everything in that dataset, then add it back to Time Machine. If problem persists,
3. Turn off AFP and delete all AFP shares, then recreate them.

If none of those work, you may have some permissions issue, which is harder to diagnose remotely.

 

[MtK]

I don't know what the problem is, but if it were me, I would proceed as follows:

1. Turn the AFP service in FreeNAS off and back on, see if that resolves it. If not,
   
2. Remove the Time Machine destination from Time Machine preferences, delete everything in that dataset, then add it back to Time Machine. If problem persists,
3. Turn off AFP and delete all AFP shares, then recreate them.

If none of those work, you may have some permissions issue, which is harder to diagnose remotely.

thanks, already did all this several times.
after a server reboot, the TimeMachine backups finished, just to figure out the other AFP Shares are not available at all (!):

 

[adrianwi]

How have you got things set-up?

I'm using TimeMachine and FreeNAS and it's been pretty problem free, with only the odd issue where a machine has failed to backup and nobody has noticed for a period of time and when it restarted wants to do a full backup again.

I've got individual datasets for each machine, and those dataset aren't used for anything else. Machines are backing up on the LAN and wirelessly.

 

[Glorious1]

First you said Time Machine tries to back up but fails, now you are saying Time Machine backups finish?
Please post the output of "zfs list" (at least the relevant part), indicate which is the Time Machine destination, and what the permissions are for that dataset and and any parent datasets.

 

[MtK]

First you said Time Machine tries to back up but fails, now you are saying Time Machine backups finish?
Please post the output of "zfs list" (at least the relevant part), indicate which is the Time Machine destination, and what the permissions are for that dataset and and any parent datasets.

yes, I said that, cause every reboot of the server or restart of the service it acts differently :/

Code:

# zfs list | grep Home
Storage/Home                                             74.0G  1.52T  34.0K  /mnt/Storage/Home
Storage/Home/USER2                                      10.3M  1.52T  10.3M  /mnt/Storage/Home/USER2
Storage/Home/MtK                                         10.9M  1.52T  10.9M  /mnt/Storage/Home/MtK
Storage/Home/TimeMachine                                 73.9G  1.52T  73.9G  /mnt/Storage/Home/TimeMachine
Storage/Home/Shared                                     10.9M  1.52T  10.9M  /mnt/Storage/Home/Shared

 

[Glorious1]

I think one problem may be that you are sharing a dataset (Home) and then sharing the children of that dataset (including TimeMachine). I'm not sure that works well. And because you've cropped the window I can't tell who the owners are or other settings.

I suggest that, as a test, you create a TimeMachine backup dataset directly under the root dataset (Storage), then see if that will work.

Also, make sure that the root dataset (Storage) has the default ownership (root:wheel) and permissions (rwxr-xr-x) and is not shared.

 

[MtK]

Home dataset is not directly shared.
it is shared through the AFP service.

TimeMachine owner is root, group is a shared group of users (MtK + User2).


btw, now everything seems to work fine, at least when I manually click "Backup Now".
automatic backups don't seem to run...

 

[Glorious1]

Home dataset is not directly shared.
it is shared through the AFP service.

I don't understand. Is there a share of the Home dataset or not? What about Storage?

Do you have two users with independent TM backups to the same dataset? That's probably not a great idea in the long run. Like Adrian, I have a separate dataset for each machine that uses TM. And I'm not sure how TM feels about not being the owner of the TM destination. One more odd thing is that your default directory permissions for TimeMachine have no execute permissions for anyone. I would enable execute across the board for directories.

The point is, you have a lot of unusual things in your setup, and it's not surprising that it is acting erratically. I suggest you set it up in a way that other people have found to work well. Then you can start experimenting with alternatives until it breaks.

 

[MtK]

I don't understand. Is there a share of the Home dataset or not? What about Storage?

no there is no direct share to the Storage and/or Home datasets.
but, there is a home directory share for each user - see attached image.

Also, make sure that the root dataset (Storage) has the default ownership (root:wheel) and permissions (rwxr-xr-x) and is not shared.

weirdly enough it was nobody:nogroup but rwxrwxrwx.

I suggest that, as a test, you create a TimeMachine backup dataset directly under the root dataset (Storage), then see if that will work.

So now I have this:

Code:

# zfs list | grep Time
Storage/TimeMachine                                      32.0K  1.49T  32.0K  /mnt/Storage/TimeMachine
Storage/TimeMachine/MtK                                  32.0K  1.49T  32.0K  /mnt/Storage/TimeMachine/MtK

and I have a direct share for MtK's TimeMachine.

 

[Glorious1]

And . . . does it work??

1. Note that your time machine dataset is still not directly under the root dataset.
   
2. Did you reset the Storage owner and permissions to default?
3. The time machine dataset itself ( /mnt/Storage/TimeMachine/Mtk ) should be owned by Mtk, shouldn't it?
4. Wouldn't you want Mac permissions for the time machine dataset (Mtk) ?

 

[MtK]

And . . . does it work??

1. Note that your time machine dataset is still not directly under the root dataset.
   
2. Did you reset the Storage owner and permissions to default?
3. The time machine dataset itself ( /mnt/Storage/TimeMachine/Mtk ) should be owned by Mtk, shouldn't it?

I rebooted my Mac and it's still backuping...

1. Storage is my root dataset, well actually pool.
2. Storage owner/group were rest to root:wheel & rwxr-xr-x
3. Storage/TimeMachine/MtK is... take a look
   

 

[Glorious1]

OK, I think the screenshot you gave in your previous msg showed Mtk owned by root:wheel.

 

[MtK]

OK, I think the screenshot you gave in your previous msg showed Mtk owned by root:wheel.

yes, that was the previous, but I changed it before starting the backup... which is getting slower and slower.

 

[MtK]

awesome!

 

[MtK]

now I need to create a dataset/share for each user right?

 

[MtK]

well, this is not really working (again):

 

[MtK]

restarting the AFP service, doesn't really help because it doesn't kill all the processes.
after manually killing the processes, TM could not find the old backup (the one I just didn't a few minutes ago)!

 

[diedrichg]

I followed the docs to the T and I haven't had any problems whatsoever. I suggest starting from zero and following the docs.

What version FreeNAS are you on?

Your individual time machine datasets should lie in the root of your volume.
e.g.
/mnt/Storage/TimeMachine1
/mnt/Storage/TimeMachine2
etc...

 

[diedrichg]

Also, when a backup is complete, you will want to set a quota for the Time Machine datasets. If you don't - they will continue to eat away at your total storage available until nothing is left. I used a WAG of the backup used * 3 as each dataset's quota.