Recovering datas when a node was joining again the cluster (with all node datas lost)

Germain Maurice germain.maurice at linkfluence.net
Mon May 17 12:04:51 EDT 2010


Le 17/05/10 15:34, Paul R a écrit :
> What should the user do to come back to the previous level of
> replication ? A forced read repair, in other words a GET with R=2 on all
> objects of all buckets ?
Yes, I wonder too what is the best thing to do after a node crash.
Eventually, i'm doing read requests on all keys of the bucket.
I found that R=1 (on all bucket keys) on the new node will adjust the 
replication level...
I wonder if R=3 or R=1, on the node to repopulate, aim to same result ?

In order to do a read repair, we have to make read requests, but it 
implies reading the stored object.
On a read repair, i assume that returning bodies is unecessary, 
especially on large objects (that i don't have).
It would be useful to provide an API operation to test the existance of 
an object without reading it...
I red REST API documentation, i didn't find this kind of operation.

Thanks.




More information about the riak-users mailing list