Let's say I have several machines in my cluster with varying amounts
of free storage and one node goes offline. Will ES move the replicas
to the machine with the greatest amount of free storage? What happens
if by moving the shards a machine will run out of storage space? I
assume it will not move the shard (which could very well bring down
another node causing a cascading failure).
There is no consideration for space (currently), just to get to a state where an even number of shards exists on each node.
On Friday, May 27, 2011 at 3:16 AM, merrellb wrote:
Let's say I have several machines in my cluster with varying amounts
of free storage and one node goes offline. Will ES move the replicas
to the machine with the greatest amount of free storage? What happens
if by moving the shards a machine will run out of storage space? I
assume it will not move the shard (which could very well bring down
another node causing a cascading failure).
So does this mean it will move a node to satisfy the requested # of
replicas even if there is not enough space on the node? I could see
this potentially leading to cascading failure as more nodes become
full, are taken off-line and require more shards to be reallocated,
filling more nodes.
On May 27, 4:14 am, Shay Banon shay.ba...@elasticsearch.com wrote:
There is no consideration for space (currently), just to get to a state where an even number of shards exists on each node.
On Friday, May 27, 2011 at 3:16 AM, merrellb wrote:
Let's say I have several machines in my cluster with varying amounts
of free storage and one node goes offline. Will ES move the replicas
to the machine with the greatest amount of free storage? What happens
if by moving the shards a machine will run out of storage space? I
assume it will not move the shard (which could very well bring down
another node causing a cascading failure).