[Gluster-devel] bit rot support for glusterfs

[shishir gowda]

On 7 January 2014 19:04, Venky Shankar <yknev.shankar at gmail.com> wrote:
>
>> 1. Depend on change-log to recompute checksum. This eliminates
>> periodic crawl of brick/volume to update the checksum.
>
>
> Consumption would be very much similar to how Geo-replication makes use of
> journals, i.e. notification after a specified time interval. If you need
> somewhat realtime notifications, that can be worked upon. For already
> existing data on a volume you could fallback to a FS crawl and use changelog
> thereafter.

Real time updates are good to have features, but I think periodic
alerts should be sufficient for now.

>
> FWIW, an history API for changelog consumption (from bricks) would be of
> much use here (given a [start-time, end-time], fetch all changelogs given
> that change-logging was _always_ enabled in that time duration, else resort
> to FS crawl).
>
>     Venky

For bit rot support, change-log should always be enabled. When turned
on a existing volume, only then should crawl be done. Does change-log
support multiple consumers at the same time with different
configurations? If not, that needs to addressed in my opinion