FreeNAS WOL: The Saga Continues...

[R.G.]

I'm good with that. What *are* the marginal costs?

For reference - my personal situation is that I started with the idea that I wanted a reliable data server, perhaps two deep in which one was a backup for the first server. That search led me to ZFS eventually. I actually had an Open Solaris server running when Oracle pulled the plug on Open Solaris, and I reasoned that the ugliness would only get worse. I found Free BSD through looking at ZFS, and FreeNAS through FreeBSD. I tinkered with FreeNAS from back in the 7.x.x days, and waited until it grew up. Somewhere around 8.x.x it got to the level of features I needed, and put in the first server. Worked well. When I got the time and money, I started the backup server, which is where I am today.

Philosophically, I'm data-centric. I've programmed in and used so many architectures, machine languages, programming languages and operating systems that I no longer care which one I'm using. There will be a manual, reference, or course of study that will let me get to what I want to get done. As a perverse example, I'm actually fond of APL for its kinkiness, not that there are many opportunities to run it on these days. I did some DSP programming in APL back before there were DSP oriented processors just to pester the professor in the DSP course.

Data integrity is the important thing. I'm currently using AMD Athlon II processors and ASUS motherboards because they are very cost effective for providing ECC memory for ZFS (and hence FreeNAS) operation. I have lost big chunks of data before in various hardware crashes, and so I know the amounts of money I'd spend to get the data back once it's lost. I view the money up front as insurance against having to do that on the back end. And in fact, FreeNAS is an appendage to getting a ZFS system done with as little "go learn a new OS" overhead as possible. It's a time saver.

I want a tool. When the system is set up and running, I want it to operate and not need tinkering and adjustment. I don't intend to spend my spare time - assuming I ever have any - doing lingering data-caresses to the setup. I'm actually building this machine a tiny building, concrete-block 50 meters from the main house. I live in the countryside, and have the space and freedoms to put it in a free-standing concrete enclosure with the systems support to keep it running. This bit of insanity comes from wanting the data to be "off site" but still under my control. A house fire will get the house computing hub and the on-line NAS, but not the backup NAS. IPMI (as I understand it currently) gives me more abilities than I really need. FreeNAS gives me just a hair more than the minimum set, which is likely to be most cost effective. I get remote shutdown with FreeNAS. Remote wakeup really makes it worth while.

I'm not adverse to spending a bit. I am adverse to spending it if I don't have to on order to meet my needs and I've thought through those needs in some detatil. Although I did refer to WOL as a toy, that was for humorous emphasis, not a philosophical statement.

S0 - what will IPMI cost me, and what will I get? Point me to IPMI motherboards and other hardware. Help me figure out how I'm endangering my data by doing shutdown by the FreeNAS "shutdown" command when needed, and re-starting by WOL. I'm sure that indiscriminate use is dangerous - but if "here be dragons", what dragons are there, and what KIND of dragons are they? I'll be happy to go do the legwork for the details, but your insight on the path to travel would be appreciated.

 

[cyberjock]

Just as either an angel's advocate or devil's advocate, don't know which for sure, when I first got into ZFS, back when Open Solaris was a possibility, I actually designed a remote power on button using a net enabled I/O adapter, which served as a remote control power on button presser. IPMI is great; I'd use it if it was more accessible and affordable. However, after a several-weeks-chase, I can now remotely power on my server, and I suspect that this small part of IPMI is enough for most people. If I had found my other post listing the string of things to check off to get it to work, I'd have had it running much sooner than I did. That's why I wrote that down.

Demeaning the idea of WOL as an inexpensive means to an end is not likely help lower the tide of incoming newbies who think "WOL - Kewl! I can do that..." I'm pretty sure that remote power on *is* what most of us are after. WOL is one way to do that. There may be even better ones, but they are hard to find.

I think that reasoned write up detailing the pitfalls of using WOL that's more detailed than "Here be dragons" and the relative advantages of IPMI, including some comparisons of the costs in money and time would be more effective at stemming the perceived tide of WOL fanatics. I personally would have loved to find that. It might even have convinced me to make the nominally "right" decision. As it is, I kept banging on it until I'd uncovered at least one path to do what I want. Pretty much, if one path is better in terms of cost, time, and danger, and the differences between paths are presented clearly, people will choose the cheaper, shorter, less dangerous path.

I used to manage both hardware and software design groups. I always found it was more effective and longer lasting to show people why doing the right thing was better than it was to tell them how silly doing the wrong thing is. But then, that may just be what worked better for me in an isolated circumstance. The wild, wild web may be different. :)

And so, @ser_rhaegar, tell me about the differences with IPMI, how much time and cost it would have saved me. I can still be taught.

IPMI is as simple as getting a board with IPMI! Plug it into your network, and bam.. you have instant remote access to your server from your network. No more monitors, keyboards, or mice. You can remotely install OSes and software with the virtual CD options built into the IPMI interface(if equipped).

S0 - what will IPMI cost me, and what will I get? Point me to IPMI motherboards and other hardware. Help me figure out how I'm endangering my data by doing shutdown by the FreeNAS "shutdown" command when needed, and re-starting by WOL. I'm sure that indiscriminate use is dangerous - but if "here be dragons", what dragons are there, and what KIND of dragons are they? I'll be happy to go do the legwork for the details, but your insight on the path to travel would be appreciated.

IPMI generally adds 2-5w to a motherboard power usage, and cost is usually $10-$30 if its built-in to the motherboard. Adding it as an extra accessory can cost as much as you want to pay.

I can't even begin to list all the potential consequences for powerdowns. The ones that come to mind is scrubs, SMART tests, maintenance that FreeNAS does(whatever that might be!), reduced wear and tear on your hard drives. Either you drink the kool-aid and realize that FreeNAS does things in the background and *needs* to do them, and that more stuff can, is, and will continue to be added with each new version, or you just accept the fact that you are making a horrible mistake that could result in data loss later. Take it for what it's worth.

Frankly, the cons of WOL have been discussed to death, so I'm not going to discuss the topic further. This isn't new for FreeNAS. This isn't new for file servers. And this shouldn't be new to anyone that's been a sysadmin. I've already done data recovery services for 2 people that would bootup their server, use it for a few hours a day, then shut it down. If you want to be stupid, feel free! I'll be glad to take your money when you need data recovery. ZFS recovery isn't cheap. The last guy that needed it paid $20,000USD for 450GB of data!

 

[R.G.]

I can't even begin to list all the potential consequences for powerdowns. The ones that come to mind is scrubs, SMART tests, maintenance that FreeNAS does(whatever that might be!), reduced wear and tear on your hard drives. Either you drink the kool-aid and realize that FreeNAS does things in the background and *needs* to do them, and that more stuff can, is, and will continue to be added with each new version, or you just accept the fact that you are making a horrible mistake that could result in data loss later. Take it for what it's worth.

Thank you. Although couched in pejorative terms, that was a very helpful thing to say for readers that may not have your insight and background.

I personally am aware of the issues with doing multiple power cycles per day on machines, and of scrubbing, reporting, and so on. My personal intentions are to have the thing be remote from my screen, actually in a separate building, and I want to be able to take it down and bring it back up sometimes without taking a walk, not to pop it up for a quick save and then turn it off. But others may actually think that they can/should cycle it multiple times per day and endanger their data.

And your advice might help them out, as you did in fact begin to list the consequences. Good on you.

 

[cyberjock]

[quote="R.G., post: 104561, member: 5523"My personal intentions are to have the thing be remote from my screen, actually in a separate building, and I want to be able to take it down and bring it back up sometimes without taking a walk, not to pop it up for a quick save and then turn it off. But others may actually think that they can/should cycle it multiple times per day and endanger their data.[/quote]

You really don't get it? Being off is being off. It doesn't matter if you do it once a day or only on weekends.

I love people who think they've got some bullet proof plan to get around some limitation with FreeNAS. Be it ECC RAM, be it RAIDZ1, be it virtualizing FreeNAS. They all figure it out sooner or later.. you can't cheat fate.

The reason's shouldn't matter to you. If Microsoft told you that doing X was dangerous for your OS, you'd not do it. Why you give the impression that the answer for "why" its stupid is beyond me. It's your risk!

Good luck!

 

[diskdiddler]

Help me figure out how I'm endangering my data by doing shutdown by the FreeNAS "shutdown" command when needed, and re-starting by WOL. I'm sure that indiscriminate use is dangerous - but if "here be dragons", what dragons are there, and what KIND of dragons are they? I'll be happy to go do the legwork for the details, but your insight on the path to travel would be appreciated.

I am also concerned by this, I would love a simple low power mode or schedule to drop a FreeNAS server into standby / hibernate or some other kind of ultra low power mode. I would rather not use a full shutdown but I figure it's the easiest method, does it not always dismount the disks properly?

What's the definitive standby / super low power mode? I'm in the precise same boat of having a regular NAS for myself and my FreeNAS is to be simply something I write data to once a week for backups.

 

[R.G.]

I am also concerned by this, I would love a simple low power mode or schedule to drop a FreeNAS server into standby / hibernate or some other kind of ultra low power mode. I would rather not use a full shutdown but I figure it's the easiest method, does it not always dismount the disks properly?

What's the definitive standby / super low power mode? I'm in the precise same boat of having a regular NAS for myself and my FreeNAS is to be simply something I write data to once a week for backups.

Shutdown does quiesce the disks properly.

I was put off by the style of the replies to my question ("reason's shouldn't matter to you") but I've never been one to sit still and shut up just because I was told to do so. I went ahead in searching out why there would be advice to never shut the thing down, and I came up with some reasons in my mind. The best reason I came up with is that ZFS does scheduled scrubs and other maintenance in the background. This can't happen when the machine is off. So leaving it off most of the time, or worse, all the time when data is not being written, prevents the kind of background processes that help keep the data secure.

A low power mode which spins down the disks would be almost as good. I did a lot of work in power consumption when I set up my box, and the bare processor/memory/cards use about 43 W. The disks pull a lot more when they start, but when they're up and spinning, it gets up to about 82W. Power use goes up again when it's doing something intense, like a scrub. I think that simply spinning the disks down would get about 90% of the benefit of being completely off.

That comes at the cost of having to spin them back up when you (or the system) needs to work with them. That costs you some time delay, but it can also simply add on/off cycles to the disk, and that may shorten their life, maybe.

There are probably other minor issues. If you use a dual-tier NAS system like I do, one "hot" one for file storage and a second to back up the first, or if you use your NAS for backing up your working machines, a good way to do that is with some form of rsynch or similar backups scheduled to run in the wee hours. Obviously, that can't happen if the NAS is off. I figure if you're smart enough to set up rsynch, you're smart enough to see the issue.

Your second post came in as I was typing this. :)

 

[diskdiddler]

I'm actually regretting the post now, having seen the post directly above mine, I don't know what to say.
Yes power cycling can be bad for hardware and software, 1000% yep definitely but it certainly shouldn't be a hard and fast rule. The whole point (??!!) of a well designed piece of software handling data redundancy is surely that the data should be recoverable fairly easily and in the case of FreeNAS one would assume it's got all kinds of checks and balances (depending on your disk configuration and raid type) to ensure the likelyhood of dataloss is completely minimal, unless multiple disks fail (less likely) or a motherboard (again, less likely?)

 

[R.G.]

Simulpost. See above.

 

[diskdiddler]

Well I certainly didn't plan to keep mine off all the time, I was looking at something which I power on maybe half the time? Maybe 1/3? I really don't need to use it a lot - maybe 7 times a week for a total of 60 minutes a time. Seems to me keeping that on for 168 hours a week is a bit excessive. I was thinking maybe spin it up every 2 days for half a day kind of thing.
EDIT: That being said, I really would like WoL to function on the HP Microserver before I can really utilise it, if it's not in as default. It's going to be tricky to setup.

 

[R.G.]

It is a tricky bit of setup, and there is room for error to creep in. You have to be SURE that the background data care-and-feeding gets to run properly and not be interrupted. The big danger is probably telling it to shut down when it's either doing or about to do some background work. But I haven't found a good way yet to tell when that will be.

Really, in the spirit of FreeNAS as a background utility, having FreeNAS manage its own power down like modern microprocessors do, seems like the best way to handle it. Having the controller and memory stay live all the time is a minor power problem; having a disk spin-up when you want to do random I/O is another issue.

This rapidly gets into the area of how you want multiple systems on a net that may or may not be running, busy, whatever when some data movement/checking is to happen. That gets complex. My next-door-neighbor works in a server farm keeping the farm spinning. There are about a dozen of them tending the servers 24/7 to keep it going.

 

[cyberjock]

You can always do remote powerons via IPMI but it won't be automated. There may be a way to script waking up a server remotely using IPMIFREE or equivalent, but that's beyond the scope of this forum as that kind of thing will depend on your IPMI hardware and if IPMIFREE or equivalent supports that.

 

[diskdiddler]

Surely there's code to deal with this? If you ask a system to shut down mid scrape / scrub or whatever it's doing, it needs to pause that. Even ghastly Windows knows to stop a defrag on shutdown?
I mean even "real" servers get shut down and moved from location to location from time to time, these things occur, this is a fact of technology.
Should it be shut on and off all the time? Probably not. To insinuate that it shutting on and off all the time is a mammoth risk, seems... I dunno, it reduces my faith in the product or implies the person suggesting it has less faith in it than I'd expect.

 

[cyberjock]

Sure the scrub will stop. But on powerup it'll start all over again.... from scratch. If you have a big pool that takes 3 hours but the machine is never on for 60 minutes you'll theoretically never be able to finish a scrub.

No, the product is fine. The problem is that the product is an appliance. That, by definition, means 99%+ uptime. Many of iX's professional customers do reboots twice a year or less.

 

[diskdiddler]

Well that's not so bad. You've got to bear in mind a large portion of users probably posting here are dummy end users not buying the physical product and dropping 200TB into a FreeNAS config. Mine is currently 3x2TB disks and 2x1TB (I was hoping to somehow join them all in to one nice big volume but should've got same sized disks)


On a completely unrelated matter, I was looking at the nightly section and saw a 9.3 area and 10 - both alpha. I did some googling for about 5/10 minutes and I'll be damned if I can find a specific list of advantages of 9.3 / 10. (yes I know it's foolish to run alpha software) - it was more of a curiosity on what's coming in the future. Couldn't see anything specific

 

[cyberjock]

doc.freenas.org usually has a page with the expected updates. I'm not sure if that is the case right now as the wikipedia page is being abandoned at some point in the future in lieu of documentation that will be included with FreeNAS (or something like that).

 

[R.G.]

You can always do remote powerons via IPMI but it won't be automated. There may be a way to script waking up a server remotely using IPMIFREE or equivalent, but that's beyond the scope of this forum as that kind of thing will depend on your IPMI hardware and if IPMIFREE or equivalent supports that.

Yep. I know.