is it okay to mirror/raid1 different hardisk model in freenas

[Azani Ramsan]

In my freenas system, I have several volumes running, one of them is consist of 2 3TB WD RED with different model (I forgot the serial, but they are different). Both are set up in a single mirror volume. I installed transmission plugin for downloading torrents, the location of the jail is also on this volume as well as the download folder location.
The issues i'm having are that 20% of my downloads always suffered from failed checksum test. "x file piece z which has been downloaded has failed its checksum test". The first thing I did was to check the smart info from each hdd, turned out no errors nor bad sectors was found. One of the hdd also have suffered from failure that it is not detected anymore and volume status changed into "degraded" status. Restarted freenas and the problematic hdd turned on and back to normal, ran another smart test and still no problems at all on both hdd.
While only running on a single hdd before I restarted, I have not received any checksum errors at all for entire day, but smart test shown no issues on the problematic drive.
So here are list of possible cause based on my knowledge (listed based on its chance of possibility) :

1. Mirroring problem between 2 different hdd models (with identical
   size)
2. Unreliable smart test
3. Transmission's problem (I have make sure to get its latest version)
4. Network problem (DNS, ISP, modem, router, etc)
5. CIFS problem

I'm using freenas 9.1 and access my files using cifs share.
Can anyone elaborate how likely each possible causes occur ? And what should I do about it ? Thanks, and also, Hi everyone, I'm new to this forum. Sorry if I posted in the wrong place

 

[Ericloewe]

As long as the two drives are similar enough (and most likely even if they aren't), it shouldn't be a problem.

 

[Azani Ramsan]

It shouldn't be a problem. no smart issues and should be okay. But still, the same checksum error. I unplugged 1 drive, no errors within a day, plug it back again and unplug the other, no errors within a day as well. Errors only occurs when both drive plugged in. Any suggestion ?

 

[cyberjock]

Uh, if you are unplugging and replugging stuff you'd better be doing scrubs. ZFS expects you to not pull the proverbial rug out from under it and be unmount and remounting disks that are out of sync with the pool.

 

[Azani Ramsan]

I did the scrubs. zpool is healty, smart test = no issues. But the checksum errors are still there. I got confused, is this the transmission's problem or hdd problem ? because the checksum error never happen when vdev only run in single hdd (without its mirror/parity)

there shouldn't be any issues when running jails in a mirrored vdev right ? Or do I need to update my freenas to latest version (currently 9.1.1) ?

btw these are the hardisk models : WD-WMC4N154**** and WD-WCC4N048****, both are WD Red 3TB

 

[no_connection]

"x file piece z which has been downloaded has failed its checksum test"

Is this during a recheck or during download?

 

[Azani Ramsan]

Mostly during downloads, mostly happen on big files (more than 1GB++).

 

[cyberjock]

Is your pool v5000? Thinking about this a little more I believe with v5000 if a disk ends up out of sync with the pool you actually have to remove it from the pool then readd it and let it resync. You may want to try this just to see if this fixes it.

 

[Azani Ramsan]

Thanks! I did what you told, remove it, reinsert, let it resilver for a moment and now I don't have any more checksum errors during these 2 days. How can this "out of sync" happen ? is it random or there is something that triggers it ? I will keep monitoring my volume and will report if this week my downloads run smoothly. Thanks again :)

--edit : sorry for bad grammar

 

[cyberjock]

Once a disk is "disconnected" from the system it's immediately out of sync with the next transaction. Since one disk will have one transaction that the other doesn't they'll be out of sync. Pretty much *any* activity at all will mean it's out of sync.

 

[Azani Ramsan]

It happens again after 5 days, checksum errors. I have no idea what made them out of sync. I'll try upgrading the freenas to the latest version this week and hopefully fix this problem.

 

[no_connection]

Just to be clear, you are referring to checksum from data read from disk?
It is not uncommon for corrupt data to be sent from users that then fails test when received.

 

[Azani Ramsan]

I don't know, how to tell ?

message : "x file piece z which has been downloaded has failed its checksum test"
from : torrent.c:3251

If it is true that this problem comes from corrupted data sent by user, then all above was a big misunderstanding which I though the volume going bad and need to be replaced or out of sync.

How do transmission handle corrupted data sent by user ? Does it still continue the downloads WITH the corrupted part, or automatically retry the part with healty part from another user ? Because movies I've downloaded that suffer from checksum errors, has minor glitch / artifacts on it (only several milisecond and only happen once or twice (depending on how many errors were encountered), but noticeable. Do I have to re-download the movie again from the beginning in this situation ?

Never for once, I downloaded anything using uTorrent (before I'm using freenas as download box and personal file server) that suffer from such problem nor that glitch I mentioned.

 

[no_connection]

Anything downloaded will be checked against a hash/checksum, the entire file is broken up into small pieces, usually 1MB but larger is starting to be common. Once a piece is downloaded it is checked and if it contains any errors it will be scrapped and downloaded again. It is possible to turn of this though (which is a bad idea).
If you receive too many broken pieces from an IP it is banned.
You can also force the entire file to be rechecked for errors and affected pieces would be redownloaded as it is then no longer 100% complete.
I have not tried Transmission on FreeNAS yet so I can't tell what settings there may be that could be wrong.

 

[cyberjock]

That sounds like a user sending you crap data and not a problem with your server losing/corrupting data.

 

[Azani Ramsan]

@no_connection
I'll look the transmission documentation for full list of parameters in the config file and see if all settings are properly set. Otherwise, I might look for another torrent downloader other than transmission and create its pbi for freenas. Thanks :)

@cyberjock
Yes, it wasn't a hardware fault appearantly. Should've just believe the smart tests from the beginning, they never lie :)

 

[cyberjock]

The thing to keep in mind with data reliability is that any storage medium that has serious flaws but isn't capable of at least identifying it's own problem is a major major defect for any manufacturer that offers warranties. If even 20% of your product is incapable of identifying these kinds of issues the RMAs and customer complaints will ruin your company. Example: OCZ

SMART isn't the end-all for everything. It definitely has some major problems and I wish it was more all-encompassing. But, generally if a disk reports SMART parameters that are fine the problem is more often than not found elsewhere. As most people know, the two most common components to fail in a computer are your PSU and your rotating platter-based hard drive. So while blaming the hard drive is super easy, choosing to accept that your disk is fine is sometimes harder to swallow. ;)