問題文
A larger cluster distributes node images from more than one place so that many nodes can be built at the same time. After the administrator changes an image, some nodes come up with the old contents. What step is being skipped?
選択肢
- The image has to be renamed after each change, because the nodes select what to install by name and will keep using a cached copy as long as the name they ask for has not changed since the previous build.
- Instead of in the image, the change should be made on each node.
- The nodes that came up with the old contents have faulty local disks, which is why they fell back to a previously installed copy instead of taking the new one from the network as the others did.
- The other places that distribute the image have to be told that the image changed, otherwise they keep handing out what they already hold.