Skip to content

(chore): refactor batch event processor to use blocking queue poll and take so as not to spin too much. #343

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Oct 23, 2019
Merged

(chore): refactor batch event processor to use blocking queue poll and take so as not to spin too much. #343

Changes from 3 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
4ecfce9
need to reset the deadline after it is hit once.
thomaszurkan-optimizely Oct 18, 2019
1b919a3
use flush interval for event queue polling
thomaszurkan-optimizely Oct 21, 2019
48f7fd8
use take if you have timed out enough
thomaszurkan-optimizely Oct 22, 2019
29e2067
refactor to only reset interval in one place.
thomaszurkan-optimizely Oct 22, 2019
c64b8d4
remove lambda after discussion with miked
thomaszurkan-optimizely Oct 22, 2019
a3fcaa5
fix log message
thomaszurkan-optimizely Oct 22, 2019
78f1cd6
check if deadline is equal or passed since we set timeout at deadline
thomaszurkan-optimizely Oct 23, 2019
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 24 additions & 3 deletions core-api/src/main/java/com/optimizely/ab/event/BatchEventProcessor.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@ public class BatchEventProcessor implements EventProcessor, AutoCloseable {
public static final String CONFIG_CLOSE_TIMEOUT = "event.processor.close.timeout";

public static final int DEFAULT_QUEUE_CAPACITY = 1000;
public static final int DEFAULT_WAIT_COUNT = 2;
public static final int DEFAULT_BATCH_SIZE = 10;
public static final long DEFAULT_BATCH_INTERVAL = TimeUnit.SECONDS.toMillis(30);
public static final long DEFAULT_TIMEOUT_INTERVAL = TimeUnit.SECONDS.toMillis(5);
Expand Down Expand Up @@ -122,26 +123,46 @@ public void flush() throws InterruptedException {
eventQueue.put(FLUSH_SIGNAL);
}

private interface QueueService {
Object get() throws InterruptedException;
}

public class EventConsumer implements Runnable {
private LinkedList<UserEvent> currentBatch = new LinkedList<>();
private long deadline = System.currentTimeMillis() + flushInterval;

@Override
public void run() {
try {
int waitCount = 0;

QueueService polling = () -> eventQueue.poll(flushInterval, TimeUnit.MILLISECONDS);
Copy link

@mjc1283 mjc1283 Oct 22, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

When polling the queue, should we block until the next deadline, instead of blocking for flushInterval?
QueueService take = () -> eventQueue.take();
QueueService using = polling;

while (true) {
if (System.currentTimeMillis() > deadline) {
logger.debug("Deadline exceeded flushing current batch.");
flush();
deadline = System.currentTimeMillis() + flushInterval;
}

Object item = eventQueue.poll(50, TimeUnit.MILLISECONDS);
Object item = using.get();
Copy link
Contributor

@mikecdavis mikecdavis Oct 22, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the QueueService adds a bit too much cognitive load then what we need in this already complex loop. Since we're basing our decision to use take() or poll(...) based on the number of times we're exited the poll with null items we can use a ternary operator inline:

long timeout = deadline - System.currentTimeMillis(); Object item = emptyCount > MAX_EMPTY_COUNT ? eventQueue.take() : eventQueue.poll(timeout, TimeUnit.MILLISECONDS);

Note I'm suggesting renaming from "wait" to "empty" since I think better captures the intention of the counter. I also used the dynamic timeout as opposed to the fix flushInterval as we discussed earlier.

if (item == null) {
logger.debug("Empty item, sleeping for 50ms.");
Thread.sleep(50);
logger.debug("Empty item after waiting flush interval. Flushing.");
flush();
Copy link
Contributor

@mikecdavis mikecdavis Oct 21, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This flush() is not necessary since it will be captured on the next while iteration. in general I'd like to keep the number of places we're calling flush() and setting the deadline to a minimum. (Ideally one place)
waitCount++;
if (waitCount > DEFAULT_WAIT_COUNT) {
using = take;
}
deadline = System.currentTimeMillis() + flushInterval;
continue;
}

waitCount = 0;
using = polling;

if (item == SHUTDOWN_SIGNAL) {
logger.info("Received shutdown signal.");
break;
Expand Down