-
Important Announcement for the TrueNAS Community.
The TrueNAS Community has now been moved. This forum has become READ-ONLY for historical purposes. Please feel free to join us on the new TrueNAS Community Forums
Non-ECC and ZFS Scrub?
- Thread starter Knowltey
- Start date
- Status
Look in the mirror! I'd consider you one of them!
jgreco
Resident Grinch
- Joined
- May 29, 2011
- Messages
- 18,680
Aiii! THAT was scary!
Hi,
I use EEC Ram (2 different manufact. though) on my HP ProLiant N54L home server setup w. FreeNAS.
Since I use ZFS, I am kinda worried, that even a single error, that the EEC does not correct/detect, could waste my entire storage pool.
Since I am not an expert, I am unsure if I should continue using ZFS; sure, an uncorrected RAM error is always a bad thing, but a filesystem that seems to turn one single bit failiure into a chain-reaction of data-doom, if the error just happens to corrupt the wrong kind of data, seems risky.
Since the my ZFS pools are my only backups (home user), should I rather switch to a different filesystem; should a ZFS only be used if secondary BU solutions are in place?
I use EEC Ram (2 different manufact. though) on my HP ProLiant N54L home server setup w. FreeNAS.
Since I use ZFS, I am kinda worried, that even a single error, that the EEC does not correct/detect, could waste my entire storage pool.
Since I am not an expert, I am unsure if I should continue using ZFS; sure, an uncorrected RAM error is always a bad thing, but a filesystem that seems to turn one single bit failiure into a chain-reaction of data-doom, if the error just happens to corrupt the wrong kind of data, seems risky.
Since the my ZFS pools are my only backups (home user), should I rather switch to a different filesystem; should a ZFS only be used if secondary BU solutions are in place?
titanic: If you are using ECC RAM you have nothing to fear. The reality of it is that if you had one error in your RAM every second you wouldn't have had a chance for it to not be detected in the estimated age of the universe(or something incredibly old).
So don't sweat it.
So don't sweat it.
jgreco
Resident Grinch
- Joined
- May 29, 2011
- Messages
- 18,680
Backups are always best practice, find a way to make sure your critical bits get backed up.
But you are fine. You are better-protected against loss with ZFS. A lot of the ECC discussion revolves around the possible snowball effect. ZFS lacks automatic tools to repair metadata damage, but a properly designed system with ECC and redundancy should never develop such damage.
But you are fine. You are better-protected against loss with ZFS. A lot of the ECC discussion revolves around the possible snowball effect. ZFS lacks automatic tools to repair metadata damage, but a properly designed system with ECC and redundancy should never develop such damage.
Wow, thx for the fast answers!
I really don't know much about RAM reliability; the INTEL pdf linked above seems to suggest though, that there are chances of uncorrectable?!? errors even with EEC.
I would not worry about such an error too much with my other (file-)systems, the chances are low (extremely low with EEC) and the consequences seem acceptable in relation (some data corruption, that might be repairable; maybe single files lost?). After reading your posts, my worries are indeed the snowball effects that you guys described, my conclusion was that EEC reduces the chances drasticly, but not the possible havoc if things still do go wrong.
Is there an easy way with FreeNas to get reports on EEC errors?
I really don't know much about RAM reliability; the INTEL pdf linked above seems to suggest though, that there are chances of uncorrectable?!? errors even with EEC.
I would not worry about such an error too much with my other (file-)systems, the chances are low (extremely low with EEC) and the consequences seem acceptable in relation (some data corruption, that might be repairable; maybe single files lost?). After reading your posts, my worries are indeed the snowball effects that you guys described, my conclusion was that EEC reduces the chances drasticly, but not the possible havoc if things still do go wrong.
Is there an easy way with FreeNas to get reports on EEC errors?
There are. There's something like 1x10^30 chance or something absurd that an error will go undetected. Remember, you are assuming that an error occurs, then that error won't be detected. Very slim odds. Banks use this technology and don't worry about errors because its just not that likely.
The reality is that ZFS has the same kind of chances of errors without being detected as errors. Check out dedup info at https://blogs.oracle.com/bonwick/entry/zfs_dedup. You have a chance of hash collision that is 1x10^77, or 0.00000000000000000000000000000000000000000000000000000000000000000000000000001. As that link says, that's 50 orders of magnitude less than an uncorrected ECC memory error! So I guess you could say that just using ZFS is more dangerous than ECC RAM! LOL!
The reality is that ZFS has the same kind of chances of errors without being detected as errors. Check out dedup info at https://blogs.oracle.com/bonwick/entry/zfs_dedup. You have a chance of hash collision that is 1x10^77, or 0.00000000000000000000000000000000000000000000000000000000000000000000000000001. As that link says, that's 50 orders of magnitude less than an uncorrected ECC memory error! So I guess you could say that just using ZFS is more dangerous than ECC RAM! LOL!
Ok, those chances are way lower than I assumed afer reading the Intel PDF! (altough I am not sure if they are referring to an undetected error in EEC rather then an detected error, that does not get handled right...).
Anyways, I was thinking about attaching a spare USB drive once a week to run a ZFS replication (once full and then snapshots) task manually. Is that possible without too much FreeNas "hacking"? And since you did not advise USB with ZFS in an older post (http://forums.freenas.org/threads/remounting-zfs-raid-1-2-x-1tb-usb-drive-array.8763/) does your opinion still stand in this case, where usb storage is only used as manualy operated, additional backup storage?
I cannot use the HDD build in the "usb external storage case" as an internal HDD since its still P-ATA :( .
EDIT: also the usb HDD is 500gb only and thus way smaller than the internally used HDDs. I'd only use it for the ZFS Datasets containing the most important data... .
Anyways, I was thinking about attaching a spare USB drive once a week to run a ZFS replication (once full and then snapshots) task manually. Is that possible without too much FreeNas "hacking"? And since you did not advise USB with ZFS in an older post (http://forums.freenas.org/threads/remounting-zfs-raid-1-2-x-1tb-usb-drive-array.8763/) does your opinion still stand in this case, where usb storage is only used as manualy operated, additional backup storage?
I cannot use the HDD build in the "usb external storage case" as an internal HDD since its still P-ATA :( .
EDIT: also the usb HDD is 500gb only and thus way smaller than the internally used HDDs. I'd only use it for the ZFS Datasets containing the most important data... .
- Status
Similar threads
