[zeromq-dev] Majordomo - losing messages ??

[Joerg Schneider-Simon]

Hi Gurus,


I've been banging my head on this one for some time now and hope one of
you may be able to point me int he right direction.

*The issue is that whenever requests are passed to the broker FAST, not
all of them make it to the (single) worker.*

If I introduce some delay between the requests (see sleep(1) in the
client code), all works just fine, but obviously, that's not acceptable

For reproduction of an issue I am experiencing, I created this
simplified version of my code:

Client:

https://pastebin.com/FsZt7ByY


Broker:

https://pastebin.com/VPjECKZW


Worker:

https://pastebin.com/yEJ8uzAL


When sending those 10 requests at once, without the delay in client.c:56
only a few of them (actual number varies) are being processed, then the
worker_sock blocks on

rc = zsock_recv(worker_sock, "sfs", &cmd, &address, &request);

Each message payload is just a few bytes, so this cannot possibly be a
RECV-buffer issue, can it?  ( I actually tried zsock_set_unbounded
(worker_sock);  with no success)


What am I missing??

Help me, Community, you are my only hope...


Joerg


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20200411/135d2f8e/attachment.htm>