Error - Resource temporarily unavailable about pyre HOT 11 CLOSED

Is this consistent behaviour? This error seems to indicate a node is gone or not responding. You didn't receive an error before?
It could also mean that the pipe between the zbeacon thread and the PyreNode thread is gone, full or deadlocked.

from pyre.

No, the error is not consistent.
Would there be any other error message if the pipe is gone, full or
deadlocked?
On Feb 23, 2015 6:35 AM, "Arnaud Loonstra" [email protected] wrote:

Is this consistent behaviour? This error seems to indicate a node is gone
or not responding. You didn't receive an error before?
It could also mean that the pipe between the zbeacon thread and the
PyreNode thread is gone, full or deadlocked.

—
Reply to this email directly or view it on GitHub
#52 (comment).

from pyre.

I'm not sure, I have experienced some deadlocks which I couldn't trace. That's the nasty thing about this asynchronous computing. I've created unittests for it in which I was able to reproduce it. It was solved by upgrading zmq. You could try that.

Though I would still like to be able to trace these errors better

from pyre.

I think I am running libzmq 4.0.4 - a further upgrade might be required?
Any numbers on what might saturate the pipes?
It could be possible that I had a lot of messages back ’n’ forth between 2 nodes. However, I did that same test multiple times but got this error only a few times.
Will this library be updated any further? Or would python bindings for Zyre replace it?

On Feb 23, 2015, at 8:10 AM, Arnaud Loonstra [email protected] wrote:

I'm not sure, I have experienced some deadlocks which I couldn't trace. That's the nasty thing about this asynchronous computing. I've created unittests for it in which I was able to reproduce it. It was solved by upgrading zmq. You could try that.

Though I would still like to be able to trace these errors better

—
Reply to this email directly or view it on GitHub #52 (comment).

from pyre.

The lib will updated further. I'm using it myself a lot. It might be superseded by the zyre bindings but these depend on libzmq, czmq and zyre. Which is quite a lot. Pyre just depends on zmq. That's why it is used often in environments where these libs are difficult to install.

The high watermark is set 1000 by default for a pipe.
https://github.com/zeromq/pyre/blob/master/pyre/zhelper.py#L15
For a zthread it's 100
https://github.com/zeromq/pyre/blob/master/pyre/zhelper.py#L43

But for the zactor I'm not sure. It's the zmq default.

I recall that when a water mark is reached you get the EAGAIN exception.

from pyre.

Do you mean this error - "zmq.error.Again: Resource temporarily unavailable" ??
That was the error I got. Is that different from the EAGAIN exception?

Also, isn't this library using zactors in place of zthreads? Or are there places where you use both?

So I take it that Pyre will be regularly upgraded - awesome :)
What exactly do you use it for now?

from pyre.

We're using both zthread and zactor. Still need to move everything to zactor. If you could try raising the HWM for zthread in your tests. See if that helps. I do not expect 100 messages to queue up but you never know.

See: https://fosdem.org/2015/schedule/event/deviot02/
Hopefully the videos will be up soon

from pyre.

Interesting!
I guess you would be the right person to ask - Why not joyn AllJoyn?
If you don't have an answer offhand, any articles you could point me to?

from pyre.

Why should I? I know there are many IoT efforts but most of them are too complex.
My initial research can be found here: http://www.z25.org/static/_rd_/zocp_init_plab/index.html

from pyre.

I was kind of curious why you suggested AllJoyn but can we close this issue anyways?

from pyre.

I was researching networks at the edge and had come across AllJoyn. But it seems a lot more centralized in its approach.
Yes, we should close this issue now :)

On Apr 10, 2015, at 3:56 AM, Arnaud Loonstra [email protected] wrote:

I was kind of curious why you suggested AllJoyn but can we close this issue anyways?

—
Reply to this email directly or view it on GitHub #52 (comment).

from pyre.