I'm using twisted with perspective broker for my distributed application. I have faced to a problem that when I try to log in more than 54 clients to my server and invoke a simple remote_call on my remote referenceable object, it receive the error : "too many open files"
But when I disconnect one client, another one can come in and get connected to the server.
Related
Being new to the RabbitMQ I was wondering how to deal with an offline target node.
As an example this scenario:
1 log recording application that stores logs to some persistent storage
N log publishing applications that want their logs to be written to the persistent storage via the log recording server.
There would be two options:
Each publishing application publishes it's log messages to it's local RabbitMQ instance and the log recording server must subscribe to each of these
The log recording application has it's local RabbitMQ instance on which each log publishing application delivers it's messages.
Option 1 would require me to reconfigure/recode/notify the recording application each time a new application appears or moves. Therefore I would think Option 2 is the right one, each new publishing application simply writes to the RabbitMQ Node of the recording application.
The only thing I am struggling with is how to deal with a situation in which the Node of the recording application is down. Do I need to build my own system to store the messages until it's back online or can I use some functionality of RabbitMQ to deal with that? I.e. could the local RabbitMQ of each of the publishing applications just receive the messages and forward them to the recording application RabbitMQ as soon as it's back online?
I found something about the Federated plugin be couldn't understand if that's the solution. Maybe I need something different or maybe I have to write my own local queueing system (which I hope I don't have to) to queue messages when the target Node is offline.
Any links to architectural examples or solutions are more than welcome.
BTW: https://groups.google.com/forum/#!topic/easynetq/nILIKSjxyMg states that you shouldn't be installing a RabbitMQ Node for each application, so maybe I should resort to something like MSQM or ZeroMQ (?)
From experience in what sounds like a similar situation, I would suggest using something other than a queue to store the messages locally, when offline.
Years ago, I built a system that had to work offline - no network connection at all - and then had to push messages through a message queue to the central server, when the laptop was brought back to the office.
I solved this by using a local database (sqlite at the time) to store my messages when the message queue was not available.
You should do something similar. Use a local database or even a plain text file or CSV file to store your messages when RabbitMQ is offline. When it reconnects, read the messages from your local file system and send them through RabbitMQ.
This is a good strategy to use, even if you do not expect RabbitMQ to go offline. Frankly, it will go offline at some point and you will have to deal with it. You should be prepared for that situation, and having a local store for your messages will help that.
...
regarding rqm node per application: bad idea. this adds a ton of complexity to your system. You want as few RabbitMQ nodes as you can get away with. Meaning, 1 per system (a system being comprised of many applications) when possible... with the exception of RabbitMQ clusters for availability - but that's another line of questions and design, entirely.
...
I did an interview with Aria Stewart about designing for failure with RabbitMQ and messaging systems, and have a small excerpt where she talks about how networks fail.
The point is, the network or RabbitMQ or something will fail and you will need a solution like a local datastore so that you can recover when RabbitMQ comes back online.
I have a two activeMQ(5.6.0) brokers. They use a shared kaha database so only one can be 'running' at once.
I have a (asp.net) webservice that puts a message on a queue, locally if I start and stop the brokers the webservice fails over correctly
when I test with the brokers on seperate machines it sometimes works but often I get "socketException: Connection reset" errors and the message is lost.
The connection string I am using is below. Note that I am aware NMS does not understand the priority backup command but I have left it there for the future.
failover:(tcp://MACHINE1:61616,tcp://MACHINE2:62616)?transport.initialReconnectDelay=1000&transport.timeout=10000&randomize=false&priorityBackup=true
How can I make my fail over between brokers fool proof?
The shared Kaha database was on a simple share. Currently activeMQ (or windows) cannot reliably get or release the lock in this configuration. The shared database must sit on a 'real' SAN so that both instances of the queue software see the database as being on a local filestore not a network location.
See this page for more info http://activemq.apache.org/shared-file-system-master-slave.html
I'm running in to an issue in an OS X app that creates multiple, persistent connections to the same host using NSURLConnection. I create a separate connection for different rooms, and it stays connected the entire time the room is open to consume a streaming API. When opening many rooms, it stops working correctly.
I created a separate sample app that creates 10 connections, and it seems to only allow 6 connections to work, and the others are queued. Does anyone know if there is a way to override this limit? I can't find it documented anywhere. The only workaround I've found is it seems to be per host name, so testing with "localhost" and "127.0.0.1" allows 6 connections per host. I uploaded a sample project with client and server here - http://cl.ly/1x3K0D1F072V3U2T0C0I.
I filed a Radar for something that seems like the same issue but on iOS. I found that you can't have more than 5 connections open at once. The connections don't have to be pointing to the same domain. Anything after that would be queued. So if you have 5 connections open to an extremely slow endpoint, any other connections will not go through.
Radar: http://openradar.appspot.com/radar?id=2542401
Apple's reply:
This is the effect of our NSURLConnection connection cache. It is expected. We expect to address this type of configuration with new API.
I asked if they could give me anymore information (does it vary? does the type of connection affect it?) and they said:
Unfortunately, we can't give details about the connection limit behavior.
User agents in general (Chrome, Firefox, Safari) use six simultaneous TCP connections per hostname, with potential one-offs.
You could break this limitation by using CFNetwork API (CFHTTPMessage).
Here is the CFNetwork Programming Guide.
https://developer.apple.com/library/mac/documentation/Networking/Conceptual/CFNetwork/Introduction/Introduction.html#//apple_ref/doc/uid/TP30001132
BTW, if you decide to use CFNetwork, you'll need to work around the proxy and authenticate.
Wish this could helped!
I am getting the following error trying to communicate from WCF -> WCF across the internet with a SSL certificate. I have 'No authentication required' checked in the WS-AT configuration on the server.
After a few seconds I'm getting the following message on my client, and trying to figure out what is wrong in my configuration (or understanding of WS-AT).
The flowed transaction could not be
unmarshaled. The following exception
occurred: The WS-AtomicTransaction
protocol service could not unmarshal
the flowed transaction. The following
exception occured: A fault reply with
code CoordinatorRegistrationFailed was
received. The fault reason follows:
The WS-AT protocol service failed to
register with its coordinator. A
connection could not be established.
Looking at the log file on the server I find a link to http://msdn.microsoft.com/en-us/library/aa702582(VS.90).aspx and the following trace message
Microsoft.Transactions.TransactionBridge.RegistrationCoordinatorFailed
Traced if the local TransactionManager is not able to
Register with its superior
TransactionManager due to the
inability to send a message.
So I'm wondering... is the server attempting to communicate with the client through a different channel to the SSL connection originally established by the client. I'm pretty sure thats what it's trying to do for me to end up with this message. The client is my laptop (currently in Starbucks) so there's no chance of any incoming connections.
So is it possible to use WS-AT transactions where only the server has an open firewall? Assuming it is possible - what do i have to do to allow for it? I'm finding very little information about necessary configuration - even in my massive Programming WCF book .
Bonus: Here's a few small tips for anyone trying to get WS-AT working :
If when trying to install the WS-AT tab you can't get it to appear - and you're running an x64 operating system - make sure you're using the x64 version of RegAsm.exe:
C:\Windows\Microsoft.NET\Framework64\v2.0.50727\RegAsm.exe /codebase wsatui.dll
If once installing WS-AT, configuring it for incoming connections and restarting MCAT you still get errors about the protocol being disabled you may have forgotton to restart IIS(!)
I am facing two issues when i try to connect to MQ which is deployed on a Remote Server from Weblogic Server(WLS) by creating a Foreign Server.
1. When I try to connect to MQ Queuemanager in Bindings mode(after importing the .Bindings file) i keep getting the below error in WLS console:
java.lang.UnsatisfiedLinkError: no mqjbnd05 in java.library.path
If i Switch the Transport to Client i keep getting:
JMSWMQ0018: Failed to connect to queue manager '' with connection mode 'Client' and host name 'localhost'. Check the queue manager is started and if running in client mode, check there is a listener running. Please see the linked exception for more information.
Has anyone seen this, and are there any performance implications which dictate the use of client over bindings and vice versa?
TIA
Finally i was able to resolve this, i had to recreate the .bindings file in the client mode, with changes to the IVTsetup.bat which is most likely present in
C:\Program Files\IBM\WebSphere MQ\java\bin, I had to run this
def qcf(psQCF) TRANSPORT(CLIENT) HOST(SMEKA) PORT(1415) CHANNEL(ps_SRV_CHANNEL) QMGR(psQM)
to generate the .bindings file.
Refer to this link for more details:
http://publib.boulder.ibm.com/infocenter/wbihelp/v6rxmx/index.jsp?topic=/com.ibm.wbia_adapters.doc/doc/peoplesoft/peopleso103.htm
Where the question states that I try to connect to MQ which is deployed on a Remote Server from Weblogic Server I assume this means that WLS and WMQ are on two different hosts. If that is the case, then a bindings mode connection (which relies on shared memory segments) won't work.
The client mode connection appears to be using a CF that is pointed to localhost rather than the IP or hostname of the WMQ server. This would work for an application on the same host as the queue manager but not when the app and QMgr are on separate servers.
As far as choosing between client and bindings mode, the answer is that if the QMgr is local use bindings. This provides highest reliability, best performance and XA transactionality. When using client mode, two-phase XA commit is not supported without the Extended Transactional Client. Per the JMS specification, there is an ambiguity that can exist if an app loses the connection during a COMMIT call. Depending on how the app handles this it's possible to end up with duplicate messages. (The JMS spec refers to these as "functionally duplicate.") This ambiguity is much less likely to occur with a bindings mode connection since there is no network latency and not even any traversal of the IP stack or interface. So use bindings mode where possible.
UPDATE:
Removed note about Extended Transactional Client being a chargeable component. As of April 24th, XTC is free of charge for all versions of WMQ on all platforms.