[
https://issues.apache.org/jira/browse/DISPATCH-1579?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17055160#comment-17055160
]
ASF GitHub Bot commented on DISPATCH-1579:
------------------------------------------
ganeshmurthy commented on pull request #693: DISPATCH-1579: open dedicated
inter-router sessions for data and control links
URL: https://github.com/apache/qpid-dispatch/pull/693#discussion_r389800442
##########
File path: src/router_node.c
##########
@@ -786,18 +786,47 @@ static int AMQP_link_attach_handler(void* context,
qd_link_t *link)
/**
- * Handler for flow events on links
+ * Handler for flow events on links. Flow updates include session window
+ * state, which needs to be checked for unblocking Q3.
*/
static int AMQP_link_flow_handler(void* context, qd_link_t *link)
{
- qd_router_t *router = (qd_router_t*) context;
- qdr_link_t *rlink = (qdr_link_t*) qd_link_get_context(link);
- pn_link_t *pnlink = qd_link_pn(link);
-
- if (!rlink)
- return 0;
+ qd_router_t *router = (qd_router_t*) context;
+ pn_link_t *pnlink = qd_link_pn(link);
+ pn_session_t *pn_ssn = pn_link_session(pnlink);
+
+ bool link_flowed = false;
+ if (pn_ssn) {
+ qd_session_t *qd_ssn = pn_session_get_context(pn_ssn);
+ if (qd_ssn && qd_session_is_q3_blocked(qd_ssn)) {
+ // Q3 blocked - have we drained enough outgoing bytes?
+ const size_t q3_lower = BUFFER_SIZE * QD_QLIMIT_Q3_LOWER;
+ if (pn_session_outgoing_bytes(pn_ssn) < q3_lower) {
+ // yes. We must now unblock all links that have been blocked
by Q3
+ qd_link_list_t *blinks = qd_session_q3_blocked_links(qd_ssn);
+ qd_link_t *blink = DEQ_HEAD(*blinks);
+ while (blink) {
+ qd_link_q3_unblock(blink); // removes from blinks list!
+ qdr_link_t *qdr_link = (qdr_link_t *)
qd_link_get_context(blink);
+ if (qdr_link) {
+ pn_link_t *pn_link = qd_link_pn(blink);
+ // signalling flow to the core causes the link to be
activated
+ qdr_link_flow(router->router_core, qdr_link,
pn_link_remote_credit(pn_link), pn_link_get_drain(pn_link));
+ if (blink == link)
+ link_flowed = true; // original link flowed - do
not flow again
+ }
+ blink = DEQ_HEAD(*blinks);
+ }
+ }
+ }
+ }
- qdr_link_flow(router->router_core, rlink, pn_link_remote_credit(pnlink),
pn_link_get_drain(pnlink));
+ if (!link_flowed) {
Review comment:
Can we please invert the logic here to first send out the flow on the
pertinent link and then do the loop after this is done
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Traffic hangs when multiple large messages are sent using different priorities
> ------------------------------------------------------------------------------
>
> Key: DISPATCH-1579
> URL: https://issues.apache.org/jira/browse/DISPATCH-1579
> Project: Qpid Dispatch
> Issue Type: Bug
> Components: Router Node
> Affects Versions: 1.10.0
> Reporter: Ken Giusti
> Assignee: Ken Giusti
> Priority: Blocker
> Fix For: 1.11.0
>
>
> Create a two hop router network. Attach a single receiver on one router.
> Create 10 senders on the other router. Each sender should sent a large
> number of (unsettled) messages at a given priority (IOW 10 senders, one
> sender sends priority 0 messages, another sends priority 1 messages, etc).
> If the messages are large enough to trigger Q3 eventually most (all?) of the
> senders will hang stuck waiting for either settlement or deliveries to be
> forwarded. There is no recovery.
> I will post a reproducer soon.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org
For additional commands, e-mail: dev-h...@qpid.apache.org
