aparnasuresh85 commented on code in PR #2493:
URL: https://github.com/apache/solr/pull/2493#discussion_r1697023501
##########
solr/core/src/java/org/apache/solr/handler/component/HttpShardHandler.java:
##########
@@ -211,23 +216,33 @@ public ShardResponse takeCompletedIncludingErrors() {
*/
@Override
public ShardResponse takeCompletedOrError() {
- return take(true);
+ return take(true, -1);
}
- private ShardResponse take(boolean bailOnError) {
+ private ShardResponse take(boolean bailOnError, long maxAllowedTimeInMillis)
{
try {
- while (pending.get() > 0) {
- ShardResponse rsp = responses.take();
- responseFutureMap.remove(rsp);
+ long deadline = System.nanoTime();
+ if (maxAllowedTimeInMillis > 0) {
+ deadline += TimeUnit.MILLISECONDS.toNanos(maxAllowedTimeInMillis);
+ } else {
+ deadline = System.nanoTime() + TimeUnit.DAYS.toNanos(1);
+ }
+ ShardResponse previousResponse = null;
+ while (pending.get() > 0) {
+ long waitTime = deadline - System.nanoTime();
+ ShardResponse rsp = responses.poll(waitTime, TimeUnit.NANOSECONDS);
pending.decrementAndGet();
+ if (rsp == null) return previousResponse;
Review Comment:
Rather than terminating early when the first request times out, can we throw
an exception to allow LBSolrClient to attempt the query on another replica?
Returning previousResponse upon a timeout doesn't seem appropriate. Instead,
the query should be tried on all replicas for the shard.
This approach would also enhance metric tracking. By improving error
handling and reporting in the search system, we can ensure that query failures
are clearly distinguished from partial results
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
