gmane.network.zeromq.devel

Re: Two apps on the same JVM

Erwin Karbasi — 2013-05-25T07:22:59

let you load the same native library from different classloaders. You'll
need to ensure the library is loaded in a parent classloader, available to
both application classloaders. For example, in Tomcat you would achieve this
by placing the jzmq jar in 'tomcat/lib' and the native libraries in
'tomcat/bin'.
(https://github.com/zeromq/jeromq), which is a pure Java implementation with
no need for native libraries.
JZMQ.jar as part of the applications  jar.
following error:
(instead of above approach of packaging the lib as part of the application
jar) and use separate Context object per application on the same JVM but we
came across following error:
lists.zeromq.orghttp://lists.zeromq.org/mailman/listinfo/zeromq-dev


Hello Siam,

Thanks a lot for fast turnaround response.

I'm using WebSphere application server.

1. I know that if I'll add the library as a shared lib to the JVM I can use
two Context object in the same JVM, but is it possible to add the
LD_LIBERARY_PATH to the machine properties or using -Djava

labview-zmq dev seems to be burned out.

crocket — 2013-05-24T21:40:20

Is there anyone to help him?

Probably donate to him?
_______________________________________________
zeromq-dev mailing list
zeromq-dev< at >lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Re: Two apps on the same JVM

Siam Rafiee — 2013-05-24T16:21:01

What container are you using to bring up the applications? The JVM won't
let you load the same native library from different classloaders. You'll
need to ensure the library is loaded in a parent classloader, available to
both application classloaders. For example, in Tomcat you would achieve
this by placing the jzmq jar in 'tomcat/lib' and the native libraries in
'tomcat/bin'.

You might find it easier to swap out JZMQ for JeroMQ (
https://github.com/zeromq/jeromq), which is a pure Java implementation with
no need for native libraries.

Siam


On 24 May 2013 12:51, Erwin Karbasi <erwin< at >optinity.com> wrote:

_______________________________________________
zeromq-dev mailing list
zeromq-dev< at >lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Re: zeromq build in the HP-UX question

AJ Lewis — 2013-05-24T16:19:04

Haha!  I *knew* there was a good reason for that one! :D  I thought
it was just mulithread enabling. Good to know!

Re: zeromq build in the HP-UX question

Ivan Pechorin — 2013-05-24T16:13:27

2013/5/24 AJ Lewis <aj.lewis< at >quantum.com>:

HP aCC flag "-mt" adds "-D_HPUX_SOURCE" [1]
_HPUX_SOURCE adds _XOPEN_SOURCE_EXTENDED [2]
_XOPEN_SOURCE_EXTENDED is necessary for sockaddr [3]

P.S. I always have "-mt" in my CFLAGS, CXXFLAGS and LDFLAGS.

[1] http://h21007.www2.hp.com/portal/download/files/unprot/aCxx/Online_Help/options.htm#opt-mt
[2] /usr/include/sys/stdsyms.h
[3] see /usr/include/sys/socket.h

Re: zeromq build in the HP-UX question

AJ Lewis — 2013-05-24T15:44:12

Odd...all I have is "-mt +DD64" in my CFLAGS, CXXFLAGS, and LDFLAGS and it
works for me on 3.2.3 with:

HP-UX bld-hpux B.11.31 U ia64 1105973567
aCC: HP C/aC++ B3910B A.06.25

I didn't put the HPUX system together though, so maybe there's an extra
package installed that gives me those identifiers...

zmq_msg_move between processes ?

Laurent Alebarde — 2013-05-24T14:53:32

Hi all,

Is it possible to move messages between processes with zmq_msg_move, or 
does it has restrictions ? Which ones please ?

Regards,

Laurent.
_______________________________________________
zeromq-dev mailing list
zeromq-dev< at >lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Re: zeromq build in the HP-UX question

Bjorn Reese — 2013-05-24T13:28:29

[...]

Try to define the _XOPEN_SOURCE_EXTENDED macro.

Two apps on the same JVM

Erwin Karbasi — 2013-05-24T11:51:45

Hello Guys,

Following our use cases:

1. We have two applications that packaged the SO file (C lib files) and
JZMQ.jar as part of the applications jar.

We have tried to deploy these apps on the same JVM and encountered
following error:

java.lang.UnsatisfiedLinkError: jzmq (Not found in java.library.path)

2. We have also tried to set the lib files in the LD_LIBERARY_PATH (instead
of above approach of packaging the lib as part of the application jar) and
use separate Context object per application on the same JVM but we came
across following error:

java.lang.UnsatisfiedLinkError: jzmq (Library is already loaded in another
ClassLoader)

                at
java.lang.ClassLoader.loadLibraryWithPath(ClassLoader.java:1053)

                at
java.lang.ClassLoader.loadLibraryWithClassLoader(ClassLoader.java:1017)

                at java.lang.System.loadLibrary(System.java:474)

                at org.zeromq.ZMQ.<clinit>(ZMQ.java:36)

Your insight regarding above two issues would highly appreciated.

Thanks,

Re: Wikipedia entry on ZeroMQ

Pieter Hintjens — 2013-05-24T10:18:17

No problem. I'm curious whether you've seen the latest ZMTP with
security mechanisms; IMO this should solve your requirements for a
secure transport for nanomsg, and implementing ZMTP would be win-win
for nanomsg and for 0MQ users. There are a lot of projects now
implementing ZMTP.

There's an implementation of ZMTP v3.0 (NULL and PLAIN mechanisms)
already in libzmq master if you are interested.

-Pieter

On Fri, May 24, 2013 at 12:13 PM, Martin Sustrik <sustrik< at >250bpm.com> wrote:

Re: Wikipedia entry on ZeroMQ

Martin Sustrik — 2013-05-24T10:13:01

Thanks for pinging me! I'll add a note there once there's a nanomsg 
release out of the door.

Martin

Wikipedia entry on ZeroMQ

Pieter Hintjens — 2013-05-24T09:20:39

Hi all,

I've updated the Wikipedia page, which was out of date.

https://en.wikipedia.org/wiki/%C3%98MQ

Sustrik, you might want to add a reference to Nano in the History
section; I didn't do that but it should probably be there.

-Pieter

Re: Functionality of connect/disconnect bind/unbind

Siam Rafiee — 2013-05-24T09:18:38

The pending messages are destroyed because there's nowhere else for them to
go.

A socket maintains a queue per peer (each connect/bind destination being a
peer). When you do a send() on a PUSH socket, it will put the message on
the outbound queue for one of its peers (simple round robin). When you
disconnect/unbind the socket from a peer, that peer's queue is destroyed
along with all the pending messages in it. Even if ZeroMQ were to hold onto
these pending messages in a separate queue within the socket - there's
still no guarantee that you won't lose messages. Messages can be stuck in
various TCP buffers across the network stack when you fail over, out of
reach of ZeroMQ.

If you need to guarantee no messages are lost, what you really need are
application-level acks for your messages, and/or your own buffering
mechanism to resend any lost messages when you fail over.

Siam



On 23 May 2013 16:10, Trevor Bernard <trevor.bernard< at >gmail.com> wrote:

_______________________________________________
zeromq-dev m

zeromq build in the HP-UX question

zhabh — 2013-05-24T08:38:09

Hi,


the build machine is :HP-UX yysdf B.11.31 U ia64 3962423388  
the tar pkg is zeromq-3.2.3   
the compile is :aCC: HP C/aC++ B3910B A.06.22
when I exec the make command ,show the error information below,and tks very much for your help...




  CXX    libzmq_la-address.lo
"tcp_address.hpp", line 45: error #2020: identifier "sockaddr" is undefined
          tcp_address_t (const sockaddr *sa, socklen_t sa_len);
                               ^


"tcp_address.hpp", line 60: error #2020: identifier "sa_family_t" is undefined
          sa_family_t family () const;
          ^


"tcp_address.hpp", line 62: error #2020: identifier "sockaddr" is undefined
          const sockaddr *addr () const;
                ^


"tcp_address.hpp", line 72: error #2020: identifier "sockaddr" is undefined
              sockaddr generic;
              ^


"tcp_address.hpp", line 73: error #2020: identifier "sockaddr_in" is undefined
              sockaddr_in ipv4;
              ^


"tcp_address.hpp", line 74: error #2020: identi

Re: Functionality of connect/disconnect bind/unbind

Apostolis Xekoukoulotakis — 2013-05-23T19:39:18

Siam seems to have explained the reason. I doubt that there is anything you
can do.

There are 2 possible cases for this:
a) you know  beforehand that a connection will fail in which case you
preemptively stop sending msgs.
b) It happens out of the blue in which case you have already lost msgs, so
what does it matter if you lose some more.


2013/5/23 Trevor Bernard <trevor.bernard< at >gmail.com>

Re: pub-sub losing messages

Michael Keselman — 2013-05-23T16:06:37

Pieter,

It worked, thank you for the help! I was ready to give up :-)

Michael


On Thu, May 23, 2013 at 1:24 AM, Pieter Hintjens <ph< at >imatix.com> wrote:

Re: efficiency of multipart messages

Yu Dongmin — 2013-05-23T15:31:54

Hi,

Basically there's only length field overhead (1 or 8 byte) when we use multipart messages.

Sure it requires more iteration in the ZMQ core, but it should be neglect. 

So I prefer a structure that I could build a program easier to understand. 20~30 multipart messages might be more hard to program not because of performance reason.


Thanks
Min

On May 23, 2013, at 7:05 AM, Davide Faconti <faconti< at >icarustechnology.com> wrote:

Re: efficiency of multipart messages

Charles Remes — 2013-05-23T15:26:44

Do whichever method best expresses your solution. There is no speed or latency difference.

As a reminder, a multipart message is sent as an atomic unit. Until you send the final frame, the receiving side cannot see any parts.

cr

On May 23, 2013, at 9:05 AM, Davide Faconti <faconti< at >icarustechnology.com> wrote:

Re: Functionality of connect/disconnect bind/unbind

Trevor Bernard — 2013-05-23T15:10:34

Thanks for follow-up Siam.

I thought I could get away with using POLL_OUT but alas, no luck.

Is there any reason why the pending messages are destroyed?



On Thu, May 23, 2013 at 7:58 AM, Siam Rafiee <siamraf< at >gmail.com> wrote:

_______________________________________________
zeromq-dev mailing list
zeromq-dev< at >lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

efficiency of multipart messages

Davide Faconti — 2013-05-23T14:05:42

Hi,

I have a question that, probably is a bit naive.
I have a published that send complex data (an easy to serialize struct,
that is about 200 bytes, but I guess this is not relevant).

I publish this data in a non periodic way, in other words sometimes I have
to deliver 10-20 samples of this data to the subscriber. Let's use N to
refer to the number of data samples I want to send.

I have two way to do that:

1) Either I publish N messages.
2) I use multipart messages provided by ZMQ; one message the message as N
parts.

Is one of these two ways more efficient in terms of, let's say latency or
speed?

Thanks in advance.

Re: Functionality of connect/disconnect bind/unbind

Siam Rafiee — 2013-05-23T10:58:58

This behaviour actually affects my system too (looks like we're modeling
failovers in the same way), so I followed this up on the #zeromq irc
channel, where guido_g helped clarify why this is happening.

[For those more familiar with Java-style syntax, I've reproduced the code
here http://pastebin.com/BD6kL20g ].

A zmq socket maintains a separate queue for each connected peer. When you
disconnect the PUSH socket from localhost:12345, that queue is destroyed
(along with any messages pending on it). When the PUSH socket is then
connected to the new endpoint, a new (empty) queue is created. Thus, only
"msg9" (which is sent after this queue is created) will be delivered when
the PULL side eventually binds at the new endpoint.

Cheers,

Siam


On 22 May 2013 23:58, Apostolis Xekoukoulotakis <xekoukou< at >gmail.com> wrote:

_______________________________________________
zeromq-dev mailing list
zeromq-dev< at >lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev