What happens with the traffic?

  • Hi,

    as we all know the image traffic will always go to the master and will then replicated to nodes in same group.

    But what happens when master and storage node are not in the same group and image is defined to be saved in that group where the storage node belongs to? Will the traffic then go directly to the storage node?

    Regards X23

  • @Sebastian-Roth Added: https://wiki.fogproject.org/wiki/index.php?title=Location_Plugin

    Also, @Sebastian-Roth when things are hash tagged wiki, after I add them I edit the hash tag and remove the hash symbol. Just an FYI so you can do the same if you wanted.

  • Senior Developer

    Should definitely do to the wiki I rekon!

  • Senior Developer

    @x23piracy, @Quazz We discussed this in chat last night.

    Images can be assigned multiple groups now. This is all the more reason for the “Primary Group” thing people are seeing.

    The way this works.

    Multicast tasks, and Upload tasks are always performed by the “Group’s Master Node”.

    In the case an image is associated to multiple groups, without using the location plugin, these same tasks will operate, but from the Primary Group’s Master Node. (I haven’t looked over the code strongly to give absolution to this, but, if I recall when coding all this, I’m fairly sure this is how I setup the functionality when introducing the “primary group”.

    Upload taskings behave a bit differently than Download though. Particularly with Location plugin.

    Without Location
    – Find the image’s primary group if associated to other groups
    – Find the group’s master node
    – Use the master node to perform the upload.
    With Location
    – Find the location’s group master
    – Use the master node to perform the upload.

    Why doesn’t upload send to the location’s designated node if present? It will, only if that node is the master of the group. Otherwise, the reason it uses the master is because of replication. Replication scan’s all subordinate nodes and replicates the master’s image set to the subordinate nodes. If there’s an image on a node that doesn’t exist on the master, it will be deleted.

    In the case of All other imaging taskings the basics are:

    Without Location
    – Find the image’s storage groups. If multiple use the primary group.
    – Find all node’s that have the image
    – Test the nodes to get which are not being used as much as another.
    – Use the “optimal” node to perform the image tasking.
    With Location – Only group defined
    – Find all node’s that have the image within the group that’s associated with the host.
    – Test the nodes to get which are not being used as much as another.
    – Use the “optimal” node to perform the image tasking
    With Location – Node defined
    – Make sure the node has the image
    – Perform the image tasking using that node.

  • Moderator

    I believe it will send the data to whichever node is designated as master of the group the image is assigned to.

Log in to reply