"Connecting..." problem after upgrading to Workplace 1.2

[Franco Dal Molin]

Last Friday, Nov 9th, we have released Workplace 1.2. Initially, everything ran fine and looked great, but soon after (around Saturday morning) we began to have problems with user's Workplaces staying in "Connecting..." state, or loosing their connections after a short while and becoming "Offline".

We investigated these issues all weekend long and tried to track the root causes down, but unfortunately were not yet able to solve them. Of course we had performed many tests in the last weeks and months, including stress and load testing, but had not seen this specific behavior  in the past.

The problem seems to be related to the load we got on our peer-to-peer infrastructure, which is build atop the latest JXTA 2.5 open-source P2P protocols (http://jxta.dev.java.net/). Our P2P infrastructure involves so called "Rendezvous" and "Relay" super peers, which are basically sever nodes running in the background helping make the P2P network run smooth, by coordinating presence and relaying traffic. We are currently working closely with the JXTA specialists from Sun Microsystems to solve the issues.

Workplace 1.1 works fine, as it uses a completely separate super peer network. Unfortunately, there is no way to revert updated Workplace (and data) back to the old version.  Rest assured that we work with highest priority on solving this issue.

Regular situation updates will be posted here, until the problem is solved.

[Franco Dal Molin]

We found the source of the problem were able to solve it!

It was an "unfortunate" firewall security setting on our super peers that was limiting new connections per time unit. The firewall would first allow a burst of new connections to be established, but then limit the number of new connections to only one per second. This is a typical firewall configuration to prevent flooding of servers. We removed it and now everything should be fine.

The Collanos Team apologizes for any troubles this might have caused you. Thank you for your comprehension and continues support!

We will continue to perform some final tests, monitoring and optimizations in the next days, but all systems should be running fine as of now.

[mickey]

Hi Franco
unfortunately it is still not working. I try to connect from three different machines (WinXP, WinServer2003, Vista) out of 2 different LAN. No connection!
Thanks for fixing - but no stress  ;-)
Beat

[Muman]

[quote author=Franco Dal Molin link=topic=1173.msg1895#msg1895 date=1194875633]
We found the source of the problem were able to solve it![/quote]

Thanks for the heads up email Franco, it takes a while longer but it is CONNECTED now! Thanks guys, now to actually drive this thing

[Franco Dal Molin]

Unfortunately, you are indeed right. We solved only one part of the problem it seems :-(

Right now we are fighting we with socket connections on the servers. They can be established, but they can't reconnect.

We'll keep you posted!

[Muman]

I spoke too soon After updating Windows and restarting, after 10 minutes still no connection so I guess the prob remains.
Have to reinstall the prior version again to reestablish our remote work environment.

Hope you squash it real soon

[Franco Dal Molin]

Ok, it's time for another status update after a very long 17 hours day.... 

We have achieved some additional improvements! The presence system works more reliably now. We have redeployed all our super peers to run the latest Java 6 update 3. Also, we have changed some Java virtual machine server settings for optimal memory allocation for Java.

Issues that might still occur: We monitored situations when the Rendezvous lost a big portion of his edge peer connections and could not immediately re-establish them. Should this happen again, then you might suddenly find yourself "Offline" for a while, but should eventually re-connect.

Another issue is the following: We noticed that the GUI sometimes can show your presence to be "Offline", whereas in fact you are still (or again) online - and replication works! We will investigate this in the next days, as this might be a GUI related bug that was discovered.

Please share with us how the system behaves for you. Is it better now? Worse? Details of what you're noticing will helps us track problem down and refine the system further.

Unfortunately we didn't find these problems during our recent system tests, otherwise we would of course held back with the new release. On the other side, not everything can always be discovered in testing environments. How does the saying goes? "In theory, theory and practice are the same, but in practice, theory and practice differ!"

[biros]

I am a new user of Collanos and it hapend to install one day in my home computer the 1.1v and the next day 1.2 version to my office computer. I had the same problems as you discribed above. Then I unistall 1.1 and reinstall 1.2 to my home computer, and it started working (getting online)but only after some minutes, and today I have chek it again and it still needs 5 to 10 minutes till shows ONLINE.
Buy the way, don't you thing it's a good idea to have "login and password" after not using collanos for some minutes. What I mean is that if you want to leave your computer for some reason for a while, you must turn off (offline collanos) and then turn on and wait to reload and you lose time. But you have to put password (like Windos's screen saver) after some minutes that you do not use collanos, maybe it is better.
Thank any way