Github user sjcorbett commented on a diff in the pull request:
https://github.com/apache/brooklyn-server/pull/916#discussion_r158319056
--- Diff:
core/src/main/java/org/apache/brooklyn/entity/group/DynamicClusterImpl.java ---
@@ -799,6 +795,14 @@ protected Entity replaceMember(Entity member,
@Nullable Location memberLoc, Map<
/** <strong>Note</strong> for sub-classes; this method can be called
while synchronized on {@link #mutex}. */
protected Collection<Entity> grow(int delta) {
Preconditions.checkArgument(delta > 0, "Must call grow with
positive delta.");
+ Integer maxSize = config().get(MAX_SIZE);
+ if (maxSize != null) {
+ final int desiredSize = getCurrentSize() + delta;
+ if (desiredSize > maxSize) {
+ throw new Resizable.InsufficientCapacityException(
--- End diff --
I took a fairly strict interpretation of DynamicCluster's contract. If it's
not possible to fulfil a request for a particular cluster size, or *n* new
members, in full then I thought it should throw. The consumer should be
smarter, expect an `InsufficientCapacityException` and ask for less. In the
case of an auto-scaler I'd expect it to make a second attempt to resize.
If the arguments to resize and resizeByDelta are suggestions rather than
requirements then best-effort is acceptable.
---
