kbanks
05-23-2008, 09:21 AM
(reposted from a customer email)
During the test of Global Ping we noticed that nodes report back at random. We kind of expected attached node check-in first. So, the question is – how can we tell which node is local to us? This is very important, since we are going to address all nodes to execute some command, except the local. For example sleep commands… May be you have this resolved some other way?
Of course they are reporting back at random - that is the whole point of the
final optional argument that specifies a "time window" to randomly reply within
Refer again to the post "More documentation on vmStat()" that I made on 05/16/2008.
As for knowing who your "directly attached node" is, the key is realizing that the directly attached node is one hop away, all other nodes are two or more hops away.
When making multicast RPC calls, one of the parameters is TTL (Time To Live). By setting this parameter to 1, it will "hop" from your PC to the directly connected node, but will not go any further. By doing a vmStat() call with a TTL of 1, only your directly connected node will answer.
As a side note, in the upcoming release of Portal we have started bolding the directly connected node to make it stand out more.
During the test of Global Ping we noticed that nodes report back at random. We kind of expected attached node check-in first. So, the question is – how can we tell which node is local to us? This is very important, since we are going to address all nodes to execute some command, except the local. For example sleep commands… May be you have this resolved some other way?
Of course they are reporting back at random - that is the whole point of the
final optional argument that specifies a "time window" to randomly reply within
Refer again to the post "More documentation on vmStat()" that I made on 05/16/2008.
As for knowing who your "directly attached node" is, the key is realizing that the directly attached node is one hop away, all other nodes are two or more hops away.
When making multicast RPC calls, one of the parameters is TTL (Time To Live). By setting this parameter to 1, it will "hop" from your PC to the directly connected node, but will not go any further. By doing a vmStat() call with a TTL of 1, only your directly connected node will answer.
As a side note, in the upcoming release of Portal we have started bolding the directly connected node to make it stand out more.