How does MogileFS handle failure?

dormando dormando at rydia.net
Wed Apr 4 20:11:10 UTC 2007


You must monitor your devices yourself (with nagios, etc), and if you 
are _sure_ that disk is dead, mark the device as dead via mogadm.

Once the device is marked 'dead', MogileFS will internally clean off the 
device and re-replicate files that were on the now dead drive.

At gaiaonline we're running a 7 terabyte MogileFS installation. 8 
trackers, 33 hosts, ~45 drives. Over 25 million files tracked. Seems to 
work well enough :)

-Dormando

rlucindo wrote:
> Hello,
> 
> I didn't find in MogileFS's documentation how it deals with disk failure. If I configured MogileFS to have at least two copies of each file and a disk fail occur (and is impossible to recovery using fsck), how Reaper work to handle this situation? A device can automatically get the dead status?
> 
> Anyone here has the experience with MogileFS running on commodity hardware (big installations)?
> 
> Thanks
> 
> []'s
> 
> Lucindo
> --
> "Programming is an unnatural act" -
> Alan J. Perlis
> 
> 



More information about the mogilefs mailing list