OK - Help me setup my new pool, the RIGHT way.

[SilverJS]

Hi all,

First, apologies for the plethora of threads lately. I'm trying to really do it right, but I've come to the conclusion that it's probably best for me to start over on my main system - something I can do now that I have a backup system. (Specs in sig are upcoming : the Backup system is currently the main, and components for the Main are inbound as we speak. I'll do the switch this weekend, I expect.)

My first intended course of action was PM'ing someone like Cyberjock or dlavigne or something, and kindly ask that person to answer these questions, but then thought that might be construed selfish : there are others who potentially stand to gain from this information, so here goes.

---------------------

For a variety of reasons (ashift debacle, permissions SNAFU with 9.3 upgrade due to my chmod'ing back in the day, .system Error in library [likely due to unwise library name], etc., etc), I'll be starting my pool over once all of my hardware comes in, and I can put all of my data on the reconstructed backup box. So, I'd like to harness the experience in this forum and make sure I'm doing it RIGHT this time. This thread is about the Main box, btw - not the Backups one.

Intended uses of the box :

1. File server and repository. I basically use it for everything these days : iTunes libraries, XBMC (now Kodi, for those who didn't know) movies and TV shows to up to 4 clients, destination for all manners of files (so they're centrally available), like scans from my Canon MX722 or, most importantly, pictures/videos;
2. Transmission, via Jail;
3. XBMC database, via MySQL (again, in Jail);
4. Destination for Windows Back-Ups (via a small hack in Windows : Virtual Hard Disks, as W7 Home Premium doesn't allow back-ups to a network destination) for all 4 computers mentioned above;
5. TimeMachine for the Mac in my studio.

That's it for now, although I'm sure I'll ask more of the thing as time goes on. So, here's what I'm thinking as far as setup (and planning to use the Wizard as much as possible, to avoid my mucking things up) :

Pool name : Tank

6 Drives in RAIDZ2

1 Dataset for Media/Pictures/Files
1 Dataset for TimeMachine

Windows share for /mnt/Tank (to be able to move Torrents from their Finished directory, to whichever directory they will actually end up with)
AFP share for TimeMachine, obviously

Snapshots every hour or so, lasting 1 week;
Replication to Backups box.

Link Aggregation for both of X10's Intel NIC's, and using the same static IP as before.

Scrubs and SMART tests : As per CyberJock's schedule, found at :

https://forums.freenas.org/index.php?threads/need-help-setting-up-smart-zfs-scrubs.16182/#post-82403

Am I forgetting anything?

---------------

So that's it! Hoping for input from all, especially those who have learnt something the hard way, or thought : "Man, I wish I would've done X when setting things up....".

Cheers!

 

[SilverJS]

Nobody...??

 

[Robert Trevellyan]

OK, I'll offer what I can:
1. I found the whole AFP/TimeMachine thing to be so unreliable that I ended up buying licenses for Arq instead (set up to use a dataset as an SFTP destination).
2. Sharing the root of your entire pool via CIFS just seems wrong to me.
3. I see a lot of posts in these forums about LAGG not working the way people expect.

Sorry, I'm not experienced enough to offer anything more definitive

 

[SilverJS]

Robert,

Thanks, I appreciate the input. And believe me, it is much more valuable than you seem to think it is. =)

I haven't had any problems with my AFP TimeMachine backup process myself; I'll certainly research LAGG, cheers!

 

[enemy85]

For what i know, LAGG works just when lot of clients are accessing the nas simultaneously (>10), so for home users is totally useless

 

[Robert Trevellyan]

Yes, somewhere in the forums is something by cyberjock to the effect that anyone expecting 2-port LAGG to double their NAS read/write performance is likely to be disappointed.

When I tested the AFP/Time Machine combo, the first issue I had was abysmal write performance, but that might have been my NAS (it wasn't the one in my signature). Then I ran into the perennial "sparesebundle is already in use" problem, which is not unique to Time Machine on FreeNAS. I could always fix that by rebooting FreeNAS, but since it's such a pervasive problem, I consider it to be an inherent flaw with Time Machine over AFP and decided I didn't want to live with it long-term.

 

[ToBeFrank]

Make sure the time machine share is standalone. In other words, don't have it be a dataset under another dataset that is also shared. Set up the time machine dataset as a unix dataset, then share it as an AFP share. I have had zero issues when doing it this way.

Note that this means you cannot share /mnt/Tank because your time machine share would then be under a dataset that is shared.

 

[SilverJS]

Alright, I'll keep that in mind, thanks!

Note that this means you cannot share /mnt/Tank because your time machine share would then be under a dataset that is shared.

That's weird, because that's pretty much exactly what I got. In other words - my main Windows share points at my root (/mnt/Raidz2 in my case - don't ask about the pool name, that's for another thread...=)), and the AFP share and dataset are nested within that, without any issues. I see quick messages in the shell to the effect my studio mac is logging in via AFP every hour or so, like clockwork.

For reference, I followed this guide when setting the whole thing up :

http://www.notquitemainstream.com/t...-machine-for-multiple-macs-on-freenas-9-2-1-3

Regarding having access to the root via CIFS : it is definitely one of the things I wasn't crazy about. Something doesn't "feel" right about having full access via Windows - but like I said, I'm not sure how to work it so I can move any files downloaded by Transmission.

 

[ToBeFrank]

I think I followed that same guide. Every couple of weeks, time machine would pop up a window complaining that the backup is bad, and it needed to start over. I never had that problem with a time capsule so I figured it had to be the freenas. I moved the time machine dataset so it is not under an already shared dataset, and I haven't had a problem since. This could all be anecdotal, of course.

 

[Glorious1]

I thought I read somewhere on here that you shouldn't share the root dataset of a pool - only datasets that you create.

And by the way, I've got two macs backing up via Time Machine to FreeNAS hourly, over wifi, and almost no problem. Once one of the backups went bad, Time Machine said it had to be recreated, but that seems to have been a one-shot deal.

 

[ser_rhaegar]

I thought I read somewhere on here that you shouldn't share the root dataset of a pool - only datasets that you create.

And by the way, I've got two macs backing up via Time Machine to FreeNAS hourly, over wifi, and almost no problem. Once one of the backups went bad, Time Machine said it had to be recreated, but that seems to have been a one-shot deal.

In those rare instances, you can fall back to a previous snapshot (if you have them on) to before the TM backup became corrupted. This is what I do as they do become corrupted on occasion. I backup an iMac and a Macbook Air over wireless hourly too. I love the TM integration in FN.

 

[Robert Trevellyan]

I'm happy for those who are using Time Machine over AFP successfully, but please, at least once, Google "Time Machine sparesebundle is already in use", and if nothing else, notice that it returns more than 35 thousand results. Then think about whether you want to rely on this as a backup strategy. If nothing else, don't let it be your only backup for anything really important.

 

[Glorious1]

I got only four results, after correcting the spelling. Of course if you don't use the quotes you'll get all kinds of marginally relevant hits.

 

[Robert Trevellyan]

The error message that appears doesn't actually say "Time Machine sparsebundle is already in use". So if you want to be pedantic about it, Google "sparsebundle is already in use", which actually appears in the error message, and delivers over 5 thousand hits. But Google is actually very good at returning relevant results for a bunch of keywords (even with typos), so I stand by my original suggestion. How many pages of results did you go through before you found one that was "marginally relevant"?

Either way, I'm not trusting Time Machine over AFP with my important data.

 

[ToBeFrank]

The error message that appears doesn't actually say "Time Machine sparsebundle is already in use". So if you want to be pedantic about it, Google "sparsebundle is already in use", which actually appears in the error message, and delivers over 5 thousand hits. But Google is actually very good at returning relevant results for a bunch of keywords (even with typos), so I stand by my original suggestion. How many pages of results did you go through before you found one that was "marginally relevant"?

Either way, I'm not trusting Time Machine over AFP with my important data.

This was an issue with the airport, and it has been fixed. Personally, I've been using a time capsule since they first came out with it, and I've never seen that error. Now that I have my freenas set up correctly, I've seen no errors there, either. I've succesfully restored my laptop multiple times. I have no problems trusting a time machine backup. Of course, my data is also backed up offsite.

 

[ser_rhaegar]

The error message that appears doesn't actually say "Time Machine sparsebundle is already in use". So if you want to be pedantic about it, Google "sparsebundle is already in use", which actually appears in the error message, and delivers over 5 thousand hits. But Google is actually very good at returning relevant results for a bunch of keywords (even with typos), so I stand by my original suggestion. How many pages of results did you go through before you found one that was "marginally relevant"?

Either way, I'm not trusting Time Machine over AFP with my important data.

Snapshots will fix any corruption issues with the sparsebundle itself so long as FreeNAS doesn't tank on you (got a proper rig and pool with snapshots enabled right?). Config issues not so much but that's on the end user.

I've used this for a long time and like I said before, the couple times an issue cropped up I went back a day via snapshot and resumed trucking along. Restore tests work flawlessly as well (bi monthly, restore from live FN and from backup of FN).

 

[Robert Trevellyan]

This was an issue with the airport, and it has been fixed.

The first three pages of search results return the same issue in relation to NETGEAR ReadyNAS, WD MyCloud, QNAP NAS and Raspberry Pi, in addition to Time Capsule.

Of course, my data is also backed up offsite.

Great. I would hate to think of anyone using Time Machine over AFP as their only backup for important data.

Snapshots will fix any corruption issues with the sparsebundle itself...

Assuming the snapshot wasn't taken during a Time Machine backup (this is basically the same issue as trying to restore a VM from a snapshot).

 

[ser_rhaegar]

Assuming the snapshot wasn't taken during a Time Machine backup (this is basically the same issue as trying to restore a VM from a snapshot).

Then use the previous snapshot.

 

[Robert Trevellyan]

Then use the previous snapshot.

That's a backup strategy? What if that one is also bad, or doesn't include the last two hours of work that you just lost?

I have to say, my mind is officially boggled right now. Four of the people on this thread alone have had problems at some point with Time Machine over AFP, and three of them still trust it.

For the record, I think Time Machine is wonderful, and I frequently recommend it for onsite backup to direct-attached storage.

 

[rogerh]

Then use the previous snapshot.

Probably a good idea to make sure the Time Machine backups are not synchronised with the regular snapshots!