[zeromq-dev] Exploring 0mq 3.0 - lack of monitoring

[Chuck Remes]

On Oct 31, 2011, at 9:31 AM, Andrew Hume wrote:

> really?
> i care about monitoring.
> 
> and while i have stated here repeatedly that it is a significant pain in the arse that
> zeromq won't give me any idea of teh queue length (and please, we udnerstand
> all about the necessary imprecisions involved -- there is no earthly reason
> why an estimate couldn't be provided!), i drank the kool-aid and did app-level monitoring
> and have honestly never looked back.

This is not true for "writes." 0mq *will* give you an estimate of the queue length. If you use non-blocking writes, it will generate EAGAIN when you have hit your HWM. So when you get EAGAIN, then you "know" a rough estimate of your queue length.

On the read side, there is no way to determine queue length in-band. You would need to setup a second pair of sockets between your writers and readers so the writer could publish its queue length when it hit EAGAIN as described above. Not optimal but not terrible either.

cr