"connectivity lost, restoring connection" warning
Hi all, I was taking another look at Mango after playing with it a while back. An upcoming project calls for some historian/SCADA work with thick/thinclient HMI's and a supervisory "booth" so a cost effective web-based server solution is being investigated. Tag count won't be over 350 but 2 of the pieces of equipment need certain sensors to be logged 1/s with and charted up live. So I threw a little something together with the VM I'm working on as my data source to populate some garbage data.
Here's my issue. Using the watchlist function I made a few windows to sort of simulate thin clients and then played with the chart update functions. Going under 5/s caused a "connectivity lost, restoring connection" warning. This would eventually spread to the other open windows and then "resolve" itself. But, while I'm getting this warning and the data points are XXXX'd out, I'm still seeing data updating and the charts updating... Is this a bug? It eventually resolves if I refresh the screen.
Upping the virtual machine to 3 cores and memory to 6GB got my windows to "resolve" faster, but the dropping error continues even though my CPU isn't maxed out and I have plenty of memory overhead (I see folks are running this on a pi, I can't see how with my test). I see this as a problem unless I'm somehow pulling and/or charting these points wrong (data source polled at 1/s, points logged as "when point value changes", . The logs show no record of a datasource losing connection, I have zero warnings or errors in the logs for the time of this test.
My typical use may see as many as 3 of these charts per screen on 2-3 screens when . user is looking closely at a sensor. Now this is with the watchlist, not a custom page, so maybe there is something more intensive about using the watchlist? I've read folks doing ms logging on here so maybe I'm missing something?
Anyway, I'd like to know before I go further into developing a project on this.
CraigWeb last edited by CraigWeb
Mango can easily log and display the resolution you are looking for however there is a limit of 5000 log entries per chart, this can be raised if it's really needed but rollups would be a much better option. With regards to the bad performance.
My first question is, are you running a licenced instance or free mode? The high-performance time-series database does not work in free mode. If you are running free mode, you are welcome to request a trial licence.
Secondly upping your VM's RAM does not mean you have given the Java VM access to all that RAM. If you are running 3.7 you can go to
MA_JAVA_OPTS="$MA_JAVA_OPTS -Xms1600m -Xmx1600m"
ms will be the min ram and mx will be the max ram allocated to the JVM in mb.
Hey thanks for the quick response. I am running the free mode so yes, only the H2 database. If getting the rates I need requires the nosql database I guess I should try that trial. I'll try changing the memory allocation in start-options too. Thanks.
CraigWeb last edited by
No problem, I'm sure you will see a huge performance increase SQL databases are pretty bad at time-series data. If you are running Linux I would give the JVM 50-60% of you allocated RAM and leave the rest for the operating system.
Dedicated 50% of the ram but honestly seeing the same problems. Only one window, roll up of like 3 seconds on a display period of 5 minutes with a 1 second interval update. So what you're saying is that the only way I'm going to see 1/s is with the noSQL database?
CraigWeb last edited by
Go to administration--->modules and click register GUID, create an account on the store site if needed. once you logged in scroll down to the bottom and click trial request. Lets start there and then if the problem persists, we will have to troubleshoot more.
Sounds good. I downloaded and installed the free version. I'm guessing I should download and replace the free version with the enterprise version?