Am 08.04.2014 um 16:48 schrieb John Sanda <jsanda(a)redhat.com>:
I think it will be simpler to let the client/agent manage the data.
The server just notifies the client that it failed to store some data, and then let the
client decide what if anything it wants to do with that data.
We have to consider the memory footprint impact on the agent, as well as the cpu usage. We
have customers where the agent alone uses 2 cpu cores to do its work (*).
Also if e.g. two servers in a cluster are down (as they would in the end talk to the same
storage nodes that are already marked as down), we need to ensure that the agent is not
running around in circles offloading the failed metrics - in theory this should not happen
with the current comm service + maint mode.
*) Some of the cpu usage has improved since the version where I have seen this.