Time Machine issue with Server 5.0.15 and OS X 10.10.5

Question

Level 4

3,594 points

Time Machine issue with Server 5.0.15 and OS X 10.10.5

I thought this might help others:

I have various servers running OS X 10.10.5 and each backed up to its neighbour by Time Machine to its neighbour. Updating from Server 5.0.4 to 5.0.15 broke this setup without providing any alerts to warn me...

As far as I can tell the combination of OS X 10.10.5 and Server 5.0.15 messes up the permissions of the TM repository - it works only if you opt to allow 'all users' to use your server as a TM repository, which is not what I needed. Anything more restricted than this, I wanted just admin users to have access, causes 'backup disk not available' alerts on the client. A fix that seems to have worked in my case was to use the command line to change the group of the 'Backups' folders and the TM images they contained:

sudo chown -R _ldap_replicator:admin Backups/

The _ldap_replicator (uid 299) user what TM is using to make the connection, which seems very wrong to me.

I thought it might be a starting point for anyone whose backups are broken.

I'd be interested to know whether the same issue arises with OS X 10.11.whatever. I wouldn't normally considering upgrading to OS X10.11 this early but I do like my backups to work properly...

C.

Posted on Oct 24, 2015 12:19 PM

Reply

Answer 1

Oct 25, 2015 3:20 AM in response to cdhw

What are your symtoms @cdhw? We can do manual backups (so I suspect that our issue is not with permissions) but TM won't fire automatically since we have upgraded. We are now OS X 10.11.1 and Server 5.0.15 (build 15S4033)

Reply

Answer 2

cdhw Author

Level 4

3,594 points

Oct 25, 2015 5:57 AM in response to Trismegister

Soon after the upgrade the TM processes, which had been working faultlessly for months, seemed to be fairly comprehensively broken, e.g. backupd on one machine spent a week sending a continuous 25 MB/s to a server, a the sparse-images that held the backups all suffered from remarkably similar forms of directory corruption, etc. (Fixable with DiskWarrior but not Disk Utility BTW.) I can't be sure this was all the fault of the move to Server 5.0.15, on an enterprise system 'network errors' can be responsible for all sorts of things.

This is all stuff one plans for and has to suck up when it happens. My surprise was that resetting TM on all servers and clients and then setting it up from scratch resulted in the permissions issues described, albeit with a few typos because there is no grace period to edit questions, above.

I've seen a problem like you describe occur when the credentials in the Login Keychain are correct but the ones in the System Keychain don't match. This can happen if you have the short name, e.g. 'admin', in one and the long name 'Administrator' in the other. I think there is a bug that makes it relatively easy to end up in this state. Another possibility is that the backup volume is being left mounted so the automatic process thinks it's busy and doesn't run. Search the logs for 'backupd' to get some useful clues.

C.

Reply

Answer 3

Oct 25, 2015 5:22 PM in response to cdhw

various servers running OS X 10.10.5 and each backed up to its neighbour by Time Machine

Why would you ever want to do that? a failure in either of two machines then takes out your backup. How is this superior to having local backup drives?

Reply

Answer 4

cdhw Author

Level 4

3,594 points

Oct 25, 2015 6:10 PM in response to Grant Bennet-Alder

Each server also backs up to a locally connected Promise Pegasus 2 RAID, which I didn't mention because it wasn't relevant to the 'over a network' issue I wanted to highlight / discuss.

C.

Reply

Answer 5

Feb 28, 2016 5:44 PM in response to cdhw

I am unable to use Time Machine with Mac OS X 10.11.3 and Server 5.0.15. I have a Western Digital USB3 drive connected to my Mac mini server, and it had been functioning just fine as the TM backup device, until 1/7/2016 that is. Time Machine reports the following error: "The network backup disk does not support the required AFP features."

Reply

Answer 6

Feb 28, 2016 7:04 PM in response to Paul Johnson13

The diagnostic you are getting is one for an inadequate Network-Attached Server (NAS) as a Backup destination.

A directly-attached WD USB-3 drive should show up as a locally attached Drive, not a Network-anything.

It sound like more hardware or software is in this chain that you have let on, or your drive is VERY confused. Did you install some magical WD package?

Reply

Answer 7

Feb 29, 2016 9:54 AM in response to Grant Bennet-Alder

Grant, I agree with you that the drive should be seen as locally attached and not as a NAS drive. I am not aware of ever installing any WD software on the drive, but I will look at that. Thanks for reaffirming my position that the drive should work as connected to the Mac mini server.

Reply

Answer 8

Feb 29, 2016 9:57 AM in response to Grant Bennet-Alder

Grant, would you suggest just wiping the drive, reformatting and configuring it as the Time Machine backup device, or is that too extreme at this point?

Reply

Answer 9

Feb 29, 2016 10:12 AM in response to Paul Johnson13

Was the drive sold as an NAS? If so, the NAS functions may be in firmware in the device, not on the drive itself, and Erasing the drive may not solve the problem.

Perhaps there is a way to turn it into just a directly-connected drive that the manufacturer or the Vendor can tell you about.

I just worry that you are "working without a net" until you get your first Backup onto it as a directly-connected drive.

--------

As an aside, did you know that if you give Time Machine an Additional drive, it automatically alternates every-other drive on every-other Backup, while keeping the two independent?

Reply

Answer 10

Feb 29, 2016 11:00 AM in response to Grant Bennet-Alder

A couple years ago my WD drive used as a Time Machine backup device failed and WD sent me a USB3 drive as a replacement. It was a vanilla USB3 drive (My Book Studio). I replaced my Mac mini with a newer one that had a USB3 port, and the replacement USB3 drive functioned fine as a TM backup device until January 7, 2016, the date of my last successful backup.

Last week I did upgrade the firmware on the USB3 drive but that didn't solve the problem with TM backups. I'm pretty certain the USB3 is NOT a NAS drive.

Today I removed the drive as my TM backup device (in the Time Machine System Preference window) and then added it again. Then TM started the long process of "Preparing backup" but it reported no other backups existed, so it lost track of 6 months of backup history. I halted the backup at that point.

The good news is that time machine didn't abort with the error message I reported in my original post on this thread.

Before I proceed with backups again, maybe you have some suggestion how I can get TM to recognize the previous TM backups. I wonder if I can use Terminal to access the TM backup folders to "straighten things out". (Maybe something got corrupted.) If that's not possible I guess I can just accept the loss of the previous TM backups, and start anew with a full TM backup.

Reply

Answer 11

Feb 29, 2016 11:16 AM in response to Paul Johnson13

Time Machine backups need to be rock-solid reliable. I would never trust the backup "lost track of", even after being "fixed up", would still be adequate in case of a Big Failure.

My inclination would be to call the old ones "corrupted" and "lost" and move on. But it's your system.

Do you know about the late James Pond's wonderful site?

http://pondini.org/OSX/Home.html

.

Reply

Answer 12

Feb 29, 2016 12:01 PM in response to Grant Bennet-Alder

Thanks for the help and the link.

Reply

Answer 13

Feb 29, 2016 6:08 PM in response to Grant Bennet-Alder

I'm happy to report that Time Machine is now working. I lost no previous TM backups. Everything is there!

I followed the instructions from the pondini,org site for resetting Time Machine (basically deleting the TM Preferences). (I also deleted a lock file -- same name with a .lck extension -- but that may not have been necessary.) And then TM was able to do its thing.

Thanks again, Grant!

Reply

Answer 14

Feb 29, 2016 6:12 PM in response to Paul Johnson13

Excellent!

Glad you got it working properly again -- and with no loss of data -- that is even better!

Reply