I only know of 3 of these. OpenMedia Vault, NasForFree and FreeNas. There may be others. Some forum post mention problems on all of them. Could be that these are due to users or maybe the software isn't as reliable as it should be. I'd be interested in opinions from users.

I intend to use LSI hardware raid so one thing I'm also curious about is the ability to load additional software such as LSI's management stuff. I suspect the kernel will have no problems handling one of their older cards but wonder about BSD which I understand FreeNas uses.

I suspect I will be running the system from SSD to gain extra disk slots. Software that does lots of writes will wear these out more quickly. There is little info about on this area on any of them. I haven't much interest in statistical monitoring - just like to have a simple way of telling that a disk has failed.

I see a web interface as being essential.

Almost any comments welcome.

John
-

Hardware RAID is needed if your server firmware won't boot from a second drive when the first one fails. It is also needed if the card has battery-backed memory and the server is not on a UPS. Otherwise, you can get better performance and reliability from software RAID on the server than the slower, cheaper processor on a RAID card.

Your info on SSD is outdated. A good SSD can sustain more writes than a rotating disk these days. Imagine the wear from moving the disk heads every 10 msec. for 3 years.

NAS software is generally a nice web-based admin tool in front of basic tools:

• md RAID
• mdadm RAID management
• smartmontools disk monitoring
• LVM for disk space admin
• samba and NFS network file systems
• NUT UPS management

I've always had better luck using command line and low-level tools than GUI front ends that hide important information and never provide all of the needed options, although something like Nagios is handy for monitoring multiple systems in a browser.

This may be obvious advice, but just in case it isn't - if you're doing hardware RAID make sure to have at least one other spare RAID card of the precise same model. Otherwise, it's a nasty single point of failure. With software RAID, you can recover from a disk controller failure simply by installing the hard drives in any other computer (with sufficient SATA ports, of course).

(I don't have any opinion on dedicated NAS software. I always roll my own with Debian. I already know how to set up and administer Debian, so making and maintaining a file server just means modifying /etc/exports on top of what I do normally.)

1 members found this post helpful.

Openmedia vault of those choices. FreeNas uses zfs, which won't work with a raid controller so it's out. Really though the idea of "nas software" as mentioned just means a web gui interface. You can have far more control running a basic debian or ubuntu server installation and doing it manually. sounds harder but really once it's set up you shouldn't need to touch it apart from updates now and again.

1 members found this post helpful.

Thanks Issac. I've run various raid cards over the years often for some time usually used ones. Never had one fail but there is always a first time. I'm attracted to this approach for 2 reasons - hp microserver and it allows me to easily fit more disks by going for 2 1/2" ones. The other reason may cause arguments. I'm not convinced that advanced disk formats are the right way to go for reliability. I feel it's best to go for a simple disk format and add a ups. The Linux NAS software I mentioned sees ZFS as being the best things since slice bread. I don't for several reasons. That puts me off reflashing a raid card to change it into a sata expander. It's not entirely clear if these will support other formats. As far as ZFS goes I'll shout the old fashioned web way. NO.

I have wondered about roll my own file server. So will have a look around to see what can be done that is simple. Googling "setting up an nfs file server" looks promising. I did set my machine up as a CIFS client some time ago - had to modify and recompile the utilities as it had been crippled due to security concerns rather than fact. I'd hope this hasn't been done to NFS. In this case I suppose I could reflash a raid card and use mdadm.

A web interface is attractive as it saves having a monitor and keyboard etc plugged into the server.

I currently run mirrored raid via mdadm on my main machine and want to switch to ssd so feel I need a nas for back up. In real terms most of that could be done weekly and a small element daily or manually if say it's a photo that has just had a lot of work done on it.

John
-

I don't understand why it would be necessary to reflash a raid card. Can't you just tell the raid card to simply show the hard drives directly? I'll admit I only have direct experience with hardware RAID integrated onto motherboards, and with them it's just a matter of setting the desired RAID mode (to none) in the BIOS.

Anyway, for Debian setting up an nfs file server is as simple as:

The contents of /etc/exports looks something like:
Then, on the client, it's as simple as:
You add something like the following to /etc/fstab (replacing 10.42.0.1 with the server's IP address or name):
It's really nice and simple, although of course you can get really fancy with various options if you want or need.

I've never had that problem, but then I'm used to administering all of my computers via ssh. I take it that you're not currently used to administering your computers via text command shell.

I won't suggest you learn the command line for snobbish reasons, but rather I recommend learning command line administration for ease of use. In my experience, it's a big headache learning how to administer everything via GUI utilities. There are a lot of things which are easier and more convenient to administer via a GUI, but bread-and-butter system maintenance and configuration? Ugh...so much hunting and clicking and clicking.

With the command line, though? Even if it's something I've never done I can find it somewhere on the internet (like this forum) and then use copy-and-paste into the command shell (which could be ssh'd to another computer). To me, that's so much easier and convenient than trying to find something with screenshots and then hunting-and-clicking.

And think about the sheer effort involved in doing all those screenshots. And the fact that only a small fraction of other users out there will have even been using the same GUI utility interface. It's kind of hopeless, really, with a small handful of exceptions (such as the ubiquitous network-manager).

I think that very quickly you'll find that administration via command line (with ssh to remote into other computers) will be less effort.

Are you familiar with rsync? It's a great little utility that can copy just the changed files. Its performance is very good. So good that I honestly never use RAID1. I only use rsync to do backups. So, for example, you could have two drives in your main machine - one SSD and one spinning hard drive. You have a script to rsync from the SSD to the hard drive every couple seconds. This way, you NEVER feel any slow-down due to the slowness of the hard drive. You only ever feel the nice speed of the SSD. And if the SSD ever fails, the hard drive will be practically up-to-date.

My own preference is actually to only do rsync backups manually, rather than continuously. Why? Because my experience is that a far more likely "failure" is user error. I go "oops!" I didn't mean to delete all that!!! With RAID or continuous backup, that "oops" immediately gets duplicated on the "backup". With manual rsync, I can just copy the "oops" from the backup.

In fact, my main usage of rsync OS synchronization has to do with my RAMBOOT technique. In my case, the main OS isn't on a fast SSD. It's on an extremely fast tmpfs ramdisk. I use rsync to backup to hard drive or SSD (either local or nfs). I get super fast performance, but also backups as up-to-date as I want. This is important because tmpfs has a high risk of total failure - if the computer loses power or shuts down for any reason, that file system is GONE. It's in RAM, it just poofs out of existence! But with rsync, I get efficient incremental backups even over a network. (I use a customized initrd to load the OS from backup upon boot.)

Oh - one other thought about using rsync to synchronize things. If you have two different computers, it can actually be better to have the backup drive on the other computer - so you rsync over the network (via nfs). Note that rsync has its own networking capabilities, but you don't need to use them.

Doing it this way, you can have the backup hard drive be a practically perfect clone - including the UUIDs, mbr, grub, etc. That way, recovery in case of SSD failure is simply to power down both computers and replace the SSD with the backup hard drive.

If the sizes are different, you don't need them to be perfect clones. So long as the OS partition has the same UUID, and the mbr is pointed to it (grub), then any other partitions don't need to match up at all. You want the OS partitions to be the same size, though, so you don't end up with unexpected disk space errors when syncing them up.

I think it was noted above but really any distro can usually be easily made to perform nas function. I tend to like the BSD based dedicated nas distro's but that doesn't mean they are best for you. ZFS as well as other filesystems are being updated almost every day. ZFS isn't the fastest but it offers many features that one may wish to have. I love hardware raid but once in a while they stop making the model and you are stuck with some drives that aren't working for a while.

The raid controllers do need to be reflashed to become a SATA expander. It's often called IT mode but I'd guess that has only happened since software raid became more popular. Maybe I have been lucky with used ones I have owned in the past. One answer to the raid controller failing might be a weekly back up to a separate hard disk. Once the controller is in I have 8 sata ports available plus the 5 that are built in and one eSata. I can probably squeeze 10 in but intend to just use a 4 disk raid 5 initially and see how it works out. Rsync is the obvious thing to use.

Of late I have had mixed feelings about linux disk formats. I recently updated my machine including larger disks for /home. Previously I have always used enterprise disks and have a fair idea how long they should last. They didn't. I suspect this was down to a single power fail but also checked disk accesses out of curiosity because the replacements were noisier and I was rather surprised how often journaling etc wrote to them. Access wears disks out as well as other problems. There are now way more accesses to user home as well. Googling firefox is wearing out your ssd shows one instance of this. All in all when considering what seems to be going on in disks and in ssd's I actually wonder about going back to ext2. All the complications do is spot problems not prevent them. ZFS-extends the time taken and the number of writes.

That wont be a popular view. However raid 5 carries on if one disk fails. It uses parity corrections so why add more. SSD's do too and a number of other things. A UPS isn't as costly as it used to be. I have one drive in my machine that reports the number of parity corrected read errors. Stacks of them but no actual errors. Maybe all modern drives are doing this. It's not a large drive, 220gb and all that is on it that gets used is /tmp and /var. I run / on ssd and /home on disks so moved most system writes to a hard drive.

I've run suse / opensuse for well over 15 years so can't see myself changing. Even after all that time I am not too good in the console. Their YAST system utility can do all sorts of things quickly from the desktop without the console. Recently it did a couple of things that were problematic to some very experienced support people. Their install also lets me set up the machine as I want graphically. In the console I just search for what ever I need to do on the web and keep a note for if I want to do the same thing again. I've worked on software for an extreme length of time - don't think my brain will retain any more. It wasn't on PC's.

Roll my own does look to be favourite. There is a decent write up on setting up an NFS server on redhat's site. YAST can also do it but that aspect is missing from my version. It might do more than I want though - in other words as opensuse thinks it should be done. Ideally I wont want to have to log into it, just have it tied to my machine only.

John
-

Okay, that's cool. A headless alternative assuming YAST can't be pointed to remote systems would be running a VNC server. You use the same opensuse OS on the file server that you're used to, and run a VNC server so you can do a graphical remote into it for various tasks.

If you're getting significant amounts of spinning hard disk failures, one thing I'd look at is cooling airflow. Before I understood the importance of cooling airflow, I crammed hard drives in tight and got failures pretty often. The only good thing about that experience was that I learned the paranoid ways of doing what's necessary to ensure no hard drive is a single point of failure.

But after I got into ensuring good cooling airflow for all of my hard drives, I've had...maybe three hard drive failure in over a decade - all of them 2.5" drives (two of them were drives that had been pulled from dead laptops so it's unclear how beat up they were before).

Oh, another headless alternative - ssh X tunneling. I don't know how YAST works. Is it just one application or a suite of applications? If it's just one, then you can use ssh to remotely run it on a different computer than the X server. In other words, the application is running on your headless server, but its window(s) are displayed on a different computer.

This is actually very simple to set up (you just install ssh and, of course, X). It's simpler and less effort than doing a VNC server.

Yep. I found out about the hot disk problem a while ago. I thought I would try a small commercial nas. Inadequate cooling and disks failed far too quickly. I then bough a microserver new and very cheap some time ago but decided to fit raid into my pc instead. HP. The main airflow into the case goes past the disks on that. The microservers are a bit worrying in this respect so will try 4x 2 1/2" disk in it's standard 4x 3 1/2 inch plug in bays first. The disk carriers are plastic and look like they will block ventilation to me but I can always add holes. I also had disk heating problems a long time ago. A Taiwanese case fixed that. It had a sort of push in drive holder that could take 6 drives. Main fan mounted in front of them and it had a filter as well. Rather expensive at the time but worth it.

Also laptop disks. Well sort of. I found some 4000 odd rpm types for automotive / robotics use going cheap. Bought some spares too as the usual price is exorbitant. Apart from spin up at 5.5 amps they are pretty low power ones. That adds to the cost though as I feel it would be wise to upgrade the miserable power output of HP's supply. They take 1u units by the look of them so hopefully those will fit. May need some metal carving but the sizes are as per 1u units.

Laptop style disks may help with spin up and down wear. From experience I usually run my disks 24/7 however disks have changed a lot since I started but some are intended for 24/7 operation.

I was very naughty yesterday but it may solve a problem. The hp microserver I have is a gen 7. While nosing about I noticed that HP have a £70 cash back offer on the intro G8 at the moment. I couldn't resist it. It comes with their remote management built in. As I understand it this hooks into the vga on the server but I am not totally sure what can be done with it. I'd expect full control of the software on the machine. It has it's own network port and I do know it can be used to install. That's a help but suspect it will do a lot more. Had to rush because it had to be delivered and paid for before the 31st. I should get a reasonable price for the G7. It's still unused.

I can do a tidy really crude icewm install with opensuse running under X, looks like X anyway and should be a decent starting point. Only one problem. I've used lots of different console editors and personally think they are from the dark ages and aught to be banned but can cope with simple ones. I don't think it's even worth learning how to use some of the others. If one will scan loads of files and make changes yes but that's different.

YAST stands for Yet Another System Utility. There are all sorts of bits and pieces in it all system related. Two things I have done recently. Removed the swap partition using the partitioner. It did that, changing fstab and also installed more software. I had asked about this on their forum - looked complicated rebuild etc. It did it in seconds. We have 2 ISP's coming into the house so two separate networks. I'm not on the one with a printer on it. I thought this would have to be set up manually. With some knowledgeable help it just wouldn't work. I connect to the other network via wifi. In the end I told YAST to use DNS on it. It downloaded more software and it's worked ever since. One change it also made was to mark eth0 as the default so that's always selected for my web access. I just had to change the base ip address of my machine and then set a static address for the printer on the other router to keep cups happy. On the network side it will set up various servers and clients also vlc. Software management is also built in. It's possible to search for applications via description. There are a number of bits and pieces - some probably as complicated as using the console but others make that pointless.

Enough OpenSuse sales talk. Must admit I am also fond of their one click installs too via their software search. They are basically a KDE distro and that is what I like to run and always have. They seem to be pretty good at producing stable releases even early on which suites me.

John
-

I concur that no NAS distro can know how "you" want it set up. And there are too many options to put them all in a GUI interface. Previously it was mentioned to have a spare raid card. If you don't have a spare, and yours fails, the only way to get your data is to find an identical card from somewhere.

Not so sure about that Awesome. Say I rsync the data on the server to a disk once a week. Initially I am only aiming for about 1TB on the server so no great problem doing that. The data will all be on my machine as well. If the raid fails and I rsync from my machine to the same back up disk I have it all backed up. Find another raid card and put the data back on the server.

Some people use nas type set ups for media servers etc. I don't have any interest in that. If I need something like that we have a Dune with a large disk in it. It's not backed up at all and doesn't need to be. We mostly watch films via an amazon fire stick plus additional software. Music is a bit more of a problem but I mostly use youtube for that and files aren't that big anyway.

Really though as I am using it purely for back up I'd need the storage in my pc to fail as well to have problems. Archiving is a bit of a worry. That's just photo's. Currently my archive for those is the card that was in the camera when they were taken. I have some which are over 25 years old and the content is fine. Things have changed a lot since then though and they probably wont hold data so long reliably now. One answer might be to stick them on a disk. I have most of the sata disks I have ever used in a box and the data on them is fine years later. Disks have moved on too. It would need to be a big disk - worrying.

John
-

I don't know about you; my own personal attitude is to be most paranoid about backing up personal photos than anything else. I have more backups of that than anything else, really - right now 6 copies on 6 different computers, two of them located in different places (one computer at my parent's house, another at my work). The efficiency of rsync over the internet is such that maintaining the backups on remote systems is not a big deal.

I'm really wary of flash cards. I had one die before I had a chance to save the photos off to the computer, and that was a heartbreaking failure. Not much you can do about that, of course, other than just taking photos with (lower quality) smart phones; they can automatically backup to cloud storage quickly. My general experience with compact flash and SD cards is that they can be too flaky to really trust.

Mostly, though, my main backup strategy is simply to have two computers which have been set up as router/server with identical ip addresses, ssh keys, etc. That way, the backup server is a drop-in replacement in case of any sort of failure. Most other backups are more a matter of having sufficient extra disk space wherever. I personally don't use any RAID other than RAID0 because I'm okay with downtime. When some downtime is acceptable, this simplifies things immensely. It takes only a minute or so for me to swap main servers, not counting however long it takes for me to get home if the failure takes place when I'm away. That small downtime is a small price to pay for the vast simplifications I have not worrying about setting up and maintaining various forms of clustering, failover, etc.

But I'm also aware of various ways both of these servers could be wiped out at the same time. That's why I have off site backups.

As for all your old SATA disks ... how much actual data that you care about is really on them? For most people, the stuff that actually takes up serious amounts of space is movies/tv shows - which you've already stated you don't worry about backup for. Assuming the actual amount of data is less than half of the total capacity, you could label equal sized discs in pairs, and simply move/copy things around so each pair contains the same data. A SATA disc just sitting around not connected to anything is a pretty reliable store of data. Having two copies of should provide pretty good assurance that one copy will be good.