Recommended Posts

So, I've been using unRAID for... a long time. I love it. Works well. Cache drive, huge improvement!

 

Array is 13 disks total, including the cache drive and the parity drive, for a total of 6.7TB usable.

 

Is there any maintenance that needs to be done to it? In normal (read: NTFS) file systems, I'd think about defragging, or at least CHECKING the fragmentation; with unRAID, though, is anything needed?

 

I ran a parity check a few weeks back, and came back with 0 errors, but is anything else needed for normal maintenance?

Share this post


Link to post
Share on other sites

probably not

 

it was mentioned in this very forum that those modern (non windows *) filesystems, are less prone to fragmentation

 

in fact now with cache use, fragmentation will be minimal, as files get written on the main disks in full chunks (probably wrong choice of words) by the system

 

 

* in fact seems to be the case with windows too, as nowadays file numbers are HUMONGOUS compared to the file numbers 5 years ago, yet, a fragmentation check will routinely show very small (<5%) fragmentation, whereas >10% was usual back then

 

maybe we also changed file "use" from back then... more files edited back then, now most of us are just collectors of files we'll never see again (the worse kind of fragmentation comes from editing and enlarging already existent files)

 

 

Share this post


Link to post
Share on other sites

Great question.  I have a few ideas ... but hope others will chime in.

 

1.  Run routine parity checks (I recommend monthly).  Parity checks result in reading each and every sector of each and every disk.  The SMART monitoring systems built into each disk will detect marginal sectors and automatically remap them to "spare" sectors.  You'll never know it happened (unless you run smartctl before and after).  If a sector has gone bad, the parity check process will rewrite that sector with reconstructed data from the array.  That rewrite will also cause SMART to remap the bad sector and give you a fresh shiny new one with the right data. 

 

The other thing parity check will do is make sure your parity is correct.  unRAID does an excellent job of maintaining parity, but checking it routinely is still a good idea.

 

2.  If your computer uses air filters, make sure you clean them out or replace them periodically.  My experience is that they clog quickly and reduce airflow even when clean.  I'd rather deal with the dust entering the case - but to each his own.  You should open the case and vacuum out any accumulated dust periodically.  I normally crack open my computers at least once a year for an upgrade, and usually just vacuum them out then. 

 

3.  Monitor the heat level inside your case and on your disks.  Heat has been associated with many varied types of computer failures, so it is worth your time to invest in fans to help keep the system cool.  With the trend towards big hot drives spaced closely together, the need to have active cooling (i.e., fans) blowing air between the drives becomes even more important.  I try to keep my drives in the upper 20s, but temps into the mid to upper 30s is safe IMO.  But when temps go over 40C, I would start to get concerned.  Anything over 45C would be cause for quick action.  At 50C - I'd turn off the computer!

 

Fans can and do wear out.  They normally enter a noisy period first.  Since many people put their unRAID servers in out of the way locations, the noise may not be detected.  Make sure you check in on him periodically to make sure unusual noises aren't emanating from it. 

 

Another note on fans, ball bearing fans cost more but last longer than bushing fans.  If you are buying fans, pay a little more for the ball bearing variety.

 

4.  Smartctl is a tool that will allow you access the SMART features of your hard drives.  (See wiki article for instructions.)  SMART maintains attributes like # of remapped sectores, # sectors waiting to be remapped, high / low operating temperatures, # errors (with details on last 5), and many others.  It will also tell you an overall status and tell you if the drive thinks that failure is imminent.  Smartctl also allows you to run two different tests - a short one (takes minutes) and a long one (takes hours).  I am not sure what these tests do, and whether the long one is much different from doing a parity check, but it is something to consider doing as preventative maintenance (I may try it on my cache disk since parity check doesn't check this drive). 

 

I recommend taking and keeping the smartctl output for each drive periodically (I do it on an ad hoc basis before opening the case to do something, but probably should start doing with every routine parity check).  If the disk becomes suspect at some point in the future, having its "life story" might be helpful.

 

5.  unRAID uses a file system called reiserfs.  There is a tool called reiserfsck that will verify that the disk internal structures are consistent.  (It can also do reconstruction of you disk if you have a catastrophic failure of some type).  Although not something I do routinely, you might consider running it.  Here is a link to the wiki article that provides instructions on running reiserfsck.

 

Share this post


Link to post
Share on other sites

Really great tips... Thanks.

 

To assist in running smartctl on all your disks, below is a shell script I use...  It puts its results in the /boot/smart_results folder on the flash drive.  It automatically names the files based on the run date so you can compare earlier results to track performance

 

Joe L.

 

#!/bin/bash
SMART_RESULTS_DIR=/boot/smart_results
if [ ! -d $SMART_RESULTS_DIR ]
then
       mkdir $SMART_RESULTS_DIR
fi
dt=`date "+%Y%m%d"`
ls -l  --time-style='+%b %d %I:%M%p'  /dev/disk/by-id | grep "ata-" | grep -v "part1 ->" |
cut -d" " -f12 | cut -c7- | while read a
do
       if [ "$a" != "" ]
       then
               echo "TEST for $a on $dt" >$SMART_RESULTS_DIR/smart_${dt}_$a.txt
               smartctl -a -d ata /dev/$a >>$SMART_RESULTS_DIR/smart_${dt}_$a.txt
       fi
done
echo "Smart Control Test Results for `date` are in :"
ls $SMART_RESULTS_DIR/smart_${dt}*

 

Edited to add additional "grep" to eliminate duplicate entries for SATA devices that appear in both "ata" and "scsi" listings.

Edited to add time-format to ls command to force fixed number of output fields.

Share this post


Link to post
Share on other sites

BJP, are you sure about the numbers you're giving for heat? Even after a few minutes from a fresh start I'm in the 35-39C range, after an hour or so of transfering the big NEW drives in my case are at 50C. Then after all the transfering it goes down to the low 40s. The case has a ton of fans in it, so I know the flow is good.

Share this post


Link to post
Share on other sites

Mine are all mid 30's as well after being spun up for an hour or so.  I just finished ripping a DVD to one of the drives.  It took about 30 minutes (ripped from laptop via wireless link) and my Parity drive is 36C and the drive I was ripping to is 35C.  The server is in an air-conditioned space with air temp at 24C.  When idle for a few hours the drives all are in the upper 20's C.

 

I'd say you might be needing some more airflow unless the ambient air temperature is much higher at your server.

 

Joe L

Share this post


Link to post
Share on other sites

Mine just finished a 10+ Hour Parity Check (after 4.3.1. upgrade).. Sever sits in a Closet with door open... airflow is so-so at best...fairly cool in the office today with window open and ceiling fan going...

 

54294548pd1.jpg

Share this post


Link to post
Share on other sites

Mine is in an unfinished basement.  I haven't taken an ambient temperature down there, but it is nice and cool.

 

You need to control the airflow and force air between the drives.  They get really hot esp. on the bottom.  I've seen reports of the temp monitoring hardware mis-reporting temperature.  I've also heard that some manufacturers intentionally put the temp sensors away from the hottest drive components to underreport temps.  (Hopefully not the case for you!)

 

If I was seeing 50C (122F) and really believed it, I'd try to mess with the airflow in the case.  Maybe turning a fan around would help?  Take a look at my case (see my sig for a link).  All the fans on the front of the case are blowing air in and forcing air over the drives, all the fans on the back are blowing air out.  I actually should have more blowing out than in (creating suction inside), but I'm not having problems so haven't added more flow out the rear.

 

Are you using 80mm fans or 120mm fans? I'd recommend 120mm fans with higher cfm ratings.  They move much more air with much less noise than 80mm fans.

 

I just saw SuperW2's post and noticed disk1 was much hotter than the rest.  My guess is that it is in a place where airflow is restricted.  I would see what I could do with that one.  Otherwise they are looking pretty good.  A couple are a little warm, but nothing to worry about if this represents a worst case.

Share this post


Link to post
Share on other sites

To assist in running smartctl on all your disks, below is a shell script I use...  It puts is results in the /boot/smart_results folder on the flash drive.  It automatically names the files based on the run date so you can compare earlier results to track performance

 

Excellent.  Added here to the Best of the Forums!  (With the other smartctl links)

Share this post


Link to post
Share on other sites
1.  Run routine parity checks (I recommend monthly).

 

Brian, I like the idea of routine parity checks, but I think 'monthly' may be too often for many of us.  Although a couple of my drives are used for nightly backups and recent archiving, many of my other drives may stay spun down for months at a time.  Just to spin a drive up to run it hard for a parity check seems wrong to me, counterproductive, as almost all the wear and tear would be coming from the parity checks.  I am leaning toward a quarterly check as better for my case, and I suspect that others who use their systems similarly may also prefer less often.

 

As to temps, keep in mind that some drives report temps somewhat higher than others.  My Seagates generally report temps about 10C higher than my Samsung's, all similarly installed, similar airflow.  From my own experience and reading, I think that temps in the 50C's are not necessarily damaging, if only short term, but they would cause me to immediately check for a failing fan.  I would always prefer drive temps in the 20's (30's for my Seagates).

 

Cartierusm:  lots of airflow is good, *IF* it is getting to all the right places.  It doesn't sound like your airflow is getting between and around your drives.  You really don't need much airflow at all, to drop the temp substantially, but you need to make sure that there is a little constant flow between and all around each drive, so that there are no dead air spots.  Sometimes lots of fans work against each other, and if not positioned correctly, can leave dead air spots where heat can build up.

 

Joe:  Thanks for the script, works well, have been tweaking it since...

Share this post


Link to post
Share on other sites

Brian, I like the idea of routine parity checks, but I think 'monthly' may be too often for many of us.  Although a couple of my drives are used for nightly backups and recent archiving, many of my other drives may stay spun down for months at a time.  Just to spin a drive up to run it hard for a parity check seems wrong to me, counterproductive, as almost all the wear and tear would be coming from the parity checks.  I am leaning toward a quarterly check as better for my case, and I suspect that others who use their systems similarly may also prefer less often.

 

Rob - Your thoughts about running it quarterly make sense.  We're in th world of conjecture.  My thinking went something like this ..

 

The longer a failed drive stays in your computer, the better the chance of a "simultaneious" drive failure is.  I figure running it monthly will do a good job of making sure no drive have failed in the last month.  The wear and tear of running for 1/2 day a month isn't much IMO.  In fact, spinning it up and exercising the heads may have a positive affect on longevity.

 

Tell you what - you do it quarterly and I'll do it monthly and we'll see whose drives wear out sooner.  Guess we'll know in about 25-50 years.  ;)

Share this post


Link to post
Share on other sites

My unraid is in my attic, which has an exhuast fan to expel the hot air. It can get hot up there, but lately it's been very very warm. I think I'll just put a house fan on it with the side door of the case off and just blow lots of air on it. Do you think this will work? I only have the server on when I'm transfering moveis to it or watching a movie so it's on only a couple hours a few times a week. The fan I'm talking about will only be on when the server is on, I'll use an X10 module to turn it on.

Share this post


Link to post
Share on other sites

Computers don't like it hot.  Taking the side off the computer and blowing a house fan inside will likely help, but it does not sound like a good long term solution.  Is there an option to move it into conditioned space?

Share this post


Link to post
Share on other sites

Without some sort of special cooling (i.e. water or peltier), it is not possible to keep a running drive less than 5-10C above ambient.  So if you have a 35C attic (95F), you will be seeing 40-45C drives when they are spinning even with good cooling.  The smaller the room, the worse the situation because the unraid itself is a heat source.

 

So, those that keep their unraids in hot attics or enclosed closets are doomed.

 

However, if you are one of those that can't do much to help the ambient temp, at least make a point of running slower drives (5400RPM) to keep the temps on the south side of my "5-10C" estimate.

 

 

Bill

Share this post


Link to post
Share on other sites

My unraid is in my attic, which has an exhuast fan to expel the hot air. It can get hot up there, but lately it's been very very warm. I think I'll just put a house fan on it with the side door of the case off and just blow lots of air on it. Do you think this will work? I only have the server on when I'm transfering moveis to it or watching a movie so it's on only a couple hours a few times a week. The fan I'm talking about will only be on when the server is on, I'll use an X10 module to turn it on.

 

This is one of the reasons I purchased the Centurion 590. It has space for 4 120MM exhaust fans.

http://www.newegg.com/Product/Product.aspx?Item=N82E16811119152

 

I don't have ac in my computer room. so I need creative ways of circulating air.

This case has worked out very well for me (as long as I tape up all the open vents and force air to be sucked in through the drives).

 

over 50C for running drives is just about too hot.

My 10,000RPM drives shut off on their own at 55C.

I've seen allot of drive failures occur at the 50c range. It really does shorten life span.

Share this post


Link to post
Share on other sites

...

I've seen allot of drive failures occur at the 50c range. It really does shorten life span.

 

Not that WeeboTech ever needs backup but the massive google and parallel labs tests back up this statement. Temps in the mid 40s have been proven to be in many cases fine and in fact better than ultra cooled devices. Whilst there are still many debates about this a rule of thumb is dont freak out at 45degC but def freak out 50+

Share this post


Link to post
Share on other sites

Putting the house fan helped alot. It went from a high of 51C to around 46C at the highest but it's generally staying lower. Thanks for the help. I built myself a new executive desk, it's huge 7' long 30" deep, high gloss ebony laminate trimmed out with Paduck, which is a naturally orange colored wood. It holds my regular server and my office computer, scanner and color laser printer. So that took my regular server out of my closet in my office. So I wanted to empty the closet out all together so I decided to put my security computer and unraid in the attic to clean up a little. I'll move it if I have to but I've had computers up there before for years and nothing went wrong.

Share this post


Link to post
Share on other sites

People always say the attic is a bad place to store valuables because of the extreme temperature variances. (seasonal)

In the winter the drives will be on the cold side, In the summer the will be on the very warm to hot side.

As much as you like the clean look of having the computer in the attic, Maybe you could consider a corner in the basement?

 

Also, if the house fan helps, you should consider a better case with "filtered" side ventilation.

otherwise you are shoving dirty air into your computer.

 

Take a look at the centurion case. The front of the case has filters that you can blow out every now and then.

I'm in a ground floor hard tile apartment and you would not believe the amount of seasonal dust my computers suck in!

Share this post


Link to post
Share on other sites

Cartierusm - glad you were able to lower your temps to mid 40s, likely that will be cool enough to avoid a quick catastrphic death.  Although they may run for years (and I hope they do), my best advice would be to move the server into conditioned space where your drive temps would max out in the upper 30s or maybe low 40s.  That would likely give you better longevity and fewer problems.

 

Attic would be about my last choice for locating an unRAID server.  As bad as too hot or too cold is for drives, having large temp swings is likely the worst!  I usually don't worry too much about normal household dust - vacuuming it out once a year has never caused me any trouble. But dust from an attic - no routine cleaning and all that insulation - would have to be 10x worse.

 

I could understand putting a small computer operating as an HTPC or PVR up in an attic where it could be hidden and connected to an upstairs TV, Losing it would not be so aweful.  But to put an unRAID server with expensive disks holding valuable data in an attic - just couldn't do it.

 

Good luck!

Share this post


Link to post
Share on other sites

I appreciate all the advice, first off I don't have a basement, second off it's really only on at night when temps are cool up there, because I have an attic fan, third off it's only on at night for a couple of hours when watching a movie, never on during the day...that is until all the dvds are ripped to it. I do mass transfer, rip about 25 dvds to my regular server and then turn on the unraid and transfer everything. As far as dust it's a semi-finished attic and is actually very clean, no blow in insulation either. I have easy access to my attic so blowing out the computer won't be a problem. Again, thanks for all the help.

Share this post


Link to post
Share on other sites

I'm trying to run Joe L.'s script, but it doesn't seem to work for me.  The /boot/smart_results/ directory gets created but nothing (no log file) is saved to the new directory.  Instead I get the following:

 

Smart Control Test Results for Sun Jul  6 18:48:17 GMT +7 2008 are in :

ls: cannot access /boot/smart_results/smart_20080706*: No such file or directory

 

From what I can tell the 'do loop' part of the script is not executing for some reason. and so the log file is not being created.

 

 

Share this post


Link to post
Share on other sites

Try changing the following line in Joe L.'s script:

 

ls -l /dev/disk/by-id | grep -v "part1 ->" |

 

to

 

ls -l /dev/disk/by-id | grep -v "part1 -\>" |

Share this post


Link to post
Share on other sites

I'm afraid that is not the solution.  Biggest issue with most people will be linefeed/cr vs linefeed and spaces/punctuation/typo errors.

 

To watch as execution occurs, to debug, invoke as follows:

 

sh -xv script_name

 

So, if your script is named smarttest.sh you can invoke it as

sh -xv smarttest.sh

 

The -xv options will show variables as they are evaluated, and show lines of code as they are executed.

 

To make it much easier, attached is a more advanced version of the same script, based on conversations with bjp999 and ideas he had. unzipp and put it on your flash drive.

 

It keeps dated smart reports in subdirectories named after the serial numbers of the disks and puts a dated consolidated summary in the top level report directory.

 

Joe L.

Share this post


Link to post
Share on other sites

I was having exactly the same symptoms and this change worked for me. Will be interesting to see if this fix works for others as well.

 

Thanks for the debugging tip. That will come in very handy!

Share this post


Link to post
Share on other sites

Copyright © 2005-2017 Lime Technology, Inc. unRAID® is a registered trademark of Lime Technology, Inc.