Github user revans2 commented on a diff in the pull request:

    https://github.com/apache/storm/pull/2623#discussion_r179571966
  
    --- Diff: 
storm-server/src/main/java/org/apache/storm/scheduler/resource/strategies/scheduling/BaseResourceAwareStrategy.java
 ---
    @@ -477,45 +414,136 @@ protected String nodeToRack(RAS_Node node) {
             List<ExecutorDetails> execsScheduled = new LinkedList<>();
     
             Map<String, Queue<ExecutorDetails>> compToExecsToSchedule = new 
HashMap<>();
    -        for (Component component : componentMap.values()) {
    -            compToExecsToSchedule.put(component.getId(), new 
LinkedList<ExecutorDetails>());
    +        for (Map.Entry<String, Component> componentEntry: 
componentMap.entrySet()) {
    +            Component component = componentEntry.getValue();
    +            compToExecsToSchedule.put(component.getId(), new 
LinkedList<>());
                 for (ExecutorDetails exec : component.getExecs()) {
                     if (unassignedExecutors.contains(exec)) {
                         compToExecsToSchedule.get(component.getId()).add(exec);
    +                    LOG.info("{} has unscheduled executor {}", 
component.getId(), exec);
                     }
                 }
             }
     
    -        Set<Component> sortedComponents = sortComponents(componentMap);
    -        sortedComponents.addAll(componentMap.values());
    +        List<Component> sortedComponents = 
topologicalSortComponents(componentMap);
     
    -        for (Component currComp : sortedComponents) {
    -            Map<String, Component> neighbors = new HashMap<String, 
Component>();
    -            for (String compId : Sets.union(currComp.getChildren(), 
currComp.getParents())) {
    -                neighbors.put(compId, componentMap.get(compId));
    +        for (Component currComp: sortedComponents) {
    +            int numExecs = 
compToExecsToSchedule.get(currComp.getId()).size();
    +            for (int i = 0; i < numExecs; i++) {
    +                execsScheduled.addAll(takeExecutors(currComp, numExecs - 
i, componentMap, compToExecsToSchedule));
                 }
    -            Set<Component> sortedNeighbors = sortNeighbors(currComp, 
neighbors);
    -            Queue<ExecutorDetails> currCompExesToSched = 
compToExecsToSchedule.get(currComp.getId());
    -
    -            boolean flag = false;
    -            do {
    -                flag = false;
    -                if (!currCompExesToSched.isEmpty()) {
    -                    execsScheduled.add(currCompExesToSched.poll());
    -                    flag = true;
    -                }
    +        }
    +
    +        LOG.info("The ordering result is {}", execsScheduled);
    +
    +        return execsScheduled;
    +    }
     
    -                for (Component neighborComp : sortedNeighbors) {
    -                    Queue<ExecutorDetails> neighborCompExesToSched =
    -                        compToExecsToSchedule.get(neighborComp.getId());
    -                    if (!neighborCompExesToSched.isEmpty()) {
    -                        execsScheduled.add(neighborCompExesToSched.poll());
    -                        flag = true;
    +    private List<ExecutorDetails> takeExecutors(Component currComp, int 
numExecs,
    +                                                final Map<String, 
Component> componentMap,
    +                                                final Map<String, 
Queue<ExecutorDetails>> compToExecsToSchedule) {
    +        List<ExecutorDetails> execsScheduled = new ArrayList<>();
    +        Queue<ExecutorDetails> currQueue = 
compToExecsToSchedule.get((currComp.getId()));
    +        Set<String> sortedChildren = getSortedChildren(currComp, 
componentMap);
    +
    +        execsScheduled.add(currQueue.poll());
    +
    +        for (String childId: sortedChildren) {
    +            Component childComponent = componentMap.get(childId);
    +            Queue<ExecutorDetails> childQueue = 
compToExecsToSchedule.get(childId);
    +            int childNumExecs = childQueue.size();
    +            if (childNumExecs == 0) {
    +                continue;
    +            }
    +            int numExecsToTake = 1;
    +            if (isShuffleFromParentToChild(currComp, childComponent)) {
    +                // if it's shuffle grouping, truncate
    +                numExecsToTake = Math.max(1, childNumExecs / numExecs);
    +            } // otherwise, one-by-one
    +
    +            for (int i = 0; i < numExecsToTake; i++) {
    +                execsScheduled.addAll(takeExecutors(childComponent, 
childNumExecs, componentMap, compToExecsToSchedule));
    +            }
    +        }
    +
    +        return execsScheduled;
    +    }
    +
    +    private Set<String> getSortedChildren(Component component, final 
Map<String, Component> componentMap) {
    +        Set<String> children = component.getChildren();
    +        Set<String> sortedChildren =
    +                new TreeSet<String>((o1, o2) -> {
    +                    Component child1 = componentMap.get(o1);
    +                    Component child2 = componentMap.get(o2);
    +                    boolean child1IsShuffle = 
isShuffleFromParentToChild(component, child1);
    +                    boolean child2IsShuffle = 
isShuffleFromParentToChild(component, child2);
    +
    +                    if (child1IsShuffle && child2IsShuffle) {
    +                        return o1.compareTo(o2);
    +                    } else if (child1IsShuffle) {
    +                        return 1;
    +                    } else {
    +                        return -1;
    +                    }
    +                });
    +        sortedChildren.addAll(children);
    +        return sortedChildren;
    +    }
    +
    +    private boolean isShuffleFromParentToChild(Component parent, Component 
child) {
    --- End diff --
    
    Nit: could we rename this from `isShuffleFromParentToChild` to something 
more like `hasLocalityAwareGroupingFromParentToChild`?  I know it is longer, 
but in the future we may want to offer a way to expand this to more than just 
shuffle.


---

Reply via email to