On Mon, Oct 24, 2016 at 02:24:48PM +0530, Ravishankar N wrote:> On 10/24/2016 02:02 PM, Gandalf Corvotempesta wrote: > > Hi, > > how gluster manage a single bad sector on a disk/brick? It will kick > > out the whole brick? > Gluster does not manage sectors, it will just propagate the error > returned by the on-disk file system for that syscall to the application. > > What if the single bad sector makes a single file partially corrupted? > Ditto.The filesystem on the brick will detect the problem, and most likely aborts itself. Depending on the configuration (mount options, tune2fs) the kernel will panic or mount the filesystem in read-only mode. When the filesystem becomes read-only, the brick process will log a warning and exit. Niels -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 801 bytes Desc: not available URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20161024/ef8071a3/attachment.sig>
On 10/24/2016 02:59 PM, Niels de Vos wrote:> On Mon, Oct 24, 2016 at 02:24:48PM +0530, Ravishankar N wrote: >> On 10/24/2016 02:02 PM, Gandalf Corvotempesta wrote: >>> Hi, >>> how gluster manage a single bad sector on a disk/brick? It will kick >>> out the whole brick? >> Gluster does not manage sectors, it will just propagate the error >> returned by the on-disk file system for that syscall to the application. >>> What if the single bad sector makes a single file partially corrupted? >> Ditto. > The filesystem on the brick will detect the problem, and most likely > aborts itself. Depending on the configuration (mount options, tune2fs) > the kernel will panic or mount the filesystem in read-only mode. > > When the filesystem becomes read-only, the brick process will log a > warning and exit.Ah, I see that the health-checker thread kills the brick process if I/O on the health_check file fails. Thanks for correcting me.> > Niels
2016-10-24 11:29 GMT+02:00 Niels de Vos <ndevos at redhat.com>:> The filesystem on the brick will detect the problem, and most likely > aborts itself. Depending on the configuration (mount options, tune2fs) > the kernel will panic or mount the filesystem in read-only mode. > > When the filesystem becomes read-only, the brick process will log a > warning and exit.So, gluster is able to handle a single bad sector. Kernel puts the FS in readonly and the brick process will exit. When the brick process exist, gluster detect the brick as missing and so on........