You are not logged in.
Hi,
I was thinking to try ZFS. Doing some research, it quickly turned out that the ZFS-guys strongly recommend the use of ECC-memory to avoid data corruption.
However, here it is stated that this is not only a problem of ZFS but of all filesystems. Other sources state that ZFS in contrast to other filesystems does not have features dealing with data corruption and therefore ECC-memory is recommend.
I was now wondering what is actually true (as my laptop definitely does not have ECC-memory): In case this is an issue for every filesystem, it doesn't matter whether the ZFS-guys recommend ECC and I don't have it as every other filesystem has the same issues. In case the ZFS-guys have a reason to recommend ECC especially for ZFS, this might be an issue to consider for deploying ZFS vs. other filesystems.
Does anyone have some insight into this matter and could provide me with more info? I don't really find information regarding this question.
I'm especially interested how ext4 deals with that matter as it is the filesystem I currently use and maybe how BTRFS handles this.
Thanks for any info!
Last edited by Ovion (2014-11-21 22:35:29)
Offline
Not a Sysadmin issue, moving to Kernel and Hardware...
Offline
oh, sorry, I wasn't sure if it's Hardware-related or more sysadmin due to the filesystem-part.
Offline
Personally I have a Dell workstation and upon seeing it could use both non-ECC and ECC I decided to go with ECC as it was pretty much the same cost. I also use btrfs and enjoy the added peace of mind knowing that corruption due to memory errors is now far less likely.
You might like this discussion on the btrfs mailing list:
http://comments.gmane.org/gmane.comp.fi … trfs/31832
Basically from what I gather btrfs is slightly more resistant in that silent corruption due to memory errors is less likely. In most cases it seems btrfs should not write the corrupt data to the disk. Based on the FAQ entry you provided it seems this is not true for ZFS.
However, btrfs is considered less stable than ZFS and although I have been using it for over a year, based on what I have read I would consider it far more likely that a btrfs bug would wipe out your data rather than a random bit flip to the memory due to cosmic rays in ZFS.
Offline
Thanks for the info. Anything how ZFS performs in comparison with ext4 or so, regarding data corruption? I know that ZFS has features like checksums and so on to detect data corruption, how does this scale compared to memory corruption?
I for myself definitely have only non-ECC-memory, so I'm especially interested whether this is really a drawback for ZFS, especially compared to ext4 (which would be my alternative for the moment), and how that scales with regard to the benefits.
Offline