Hello,

We use ZeroMQ 4.1.1 with EPGM in our Java applications, using pub/sub sockets. 
Under performance tests at very heavy load, we intermittently see a slow 
consumer problem with one application. When receiving a burst of traffic from a 
fast sender, the receiver sometimes abruptly stops receiving messages. When 
that happens, the receiving app no longer sees any new messages from that 
sender until the app is restarted.

I presume that the sender is being muted when the receiver’s unprocessed 
messages hit a high-water mark. I’d like to understand how to monitor for this 
event, and how the receiving app can recover, if the receiver’s slowness was 
only temporary. Is there a better approach than using heartbeats, and closing 
and re-opening the listening (SUB) socket if the heartbeats from the PUB stop 
arriving?

I am not looking for you to fix our code. Under normal circumstances, the code 
works fine. I understand that in a slow-consumer situation, something has to 
give, and there are a few ways that the system could fail. Muting the sender 
seems like a reasonable failure mode. I’m looking for education, so that I can 
better understand the way that the library handles muting in this situation.

If anyone has time to explain it in this group, or can point to a good 
explanation elsewhere, that would be lovely and I’d be in your debt. If you’d 
prefer to contact me directly to set up a brief consulting arrangement where we 
could talk this through on the phone (or in person if you’re in NYC), please 
email me at jamie (at) otcmarkets (dot) com.

Regards,

  Jamie Stewart


________________________________

This e-mail (including any attachments) is confidential and may contain 
sensitive or legally privileged information. It is for the intended recipient 
only. Do not, directly or indirectly, copy, reproduce, print or disclose any 
part of this e-mail without the prior written consent of OTC Markets Group Inc. 
If you receive this message in error, please delete it and any attachments, and 
notify OTC Markets Group of misdelivery by return e-mail.
---------------------------------------------------------------------------------------
 This email has been scanned for email related threats and delivered safely by 
Mimecast.
 For more information please visit http://www.mimecast.com
---------------------------------------------------------------------------------------
_______________________________________________
zeromq-dev mailing list
[email protected]
https://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to