[Gluster-devel] self heal option

[Matt Paine]

Hi August


> Take a simple 2 machine afr mirror for example (spec file below). I
> have a brick on machine A doing afr to the brick on machine B and vice
> versa. If one of the bricks goes down and I create a file on the
> remaining one, is there any way to have healing done with the other
> brick comes back up?

At the moment the self heal functionality is passive. I.e. you need to 
do an open on the file to kick in self-heal. I believe the dev team are 
working on active self heal (like what you are describing).


You could try running this short script to find all your files and open 
them, which should self-heal every file if neccesary

find /mnt/glusterfs/ -type -f -exec head -c 1 {} \;

Or something along those lines anyway :)


> In my testing, I don't see any evidence of self-heal in afr even
> though several folks have it in their spec files. When I bring the
> other machine back online, files that were create while it was down
> are never created on the mirror brick.
> 
> Can someone with a more intimate knowledge of the code comment on this
> ? thanks...
> 


As I said - simply open that file and self-heal will kick in. As far as 
I am aware this is the way it works on both afr and unify xlators at the 
moment - and also I believe the devs are working on an automated way to 
do self heal if its a requirement.




Matt.