telljcl Posted June 19, 2018 Posted June 19, 2018 I can't be sure, but it sure seems that if I leave the Admin Console running (not same PC as one that runs Nodelink), I seem to get more errors / notifications in Nodelink. Even killing and then starting Nodelink on my server PC, I frequently get errors or notifications upon Nodelink startup that aren't repeatable every time - often different in fact with each kill / restart. Exiting Admin Console and killing / restarting Nodelink generally (if not always) produces "clean" starts of Nodelink. Should having the Admin Console running have any effect on Nodelink's operation ? Thanks
bpwwer Posted June 19, 2018 Posted June 19, 2018 I reported, what I think is the same issue to UDI a while back. When any node server is running, the admin console seems to open connections and leave them in a close-wait state. After a while, the number of connections sitting in the close-wait state fills up all the connection slots that Windows has and network operations on that PC pretty much stop since nothing can create connections any more. Closing the admin console, frees all those connections and the system goes back to normal. For me, on my Windows 7 VM, it seems to take around 24 hours for it to use all the connections. It seems independent of the particular node server being used. It's pretty easy to watch it happen with a program like TCPView. It will list all the network connections and their current state and count how many connections are in each state.
larryllix Posted June 19, 2018 Posted June 19, 2018 1 hour ago, bpwwer said: I reported, what I think is the same issue to UDI a while back. When any node server is running, the admin console seems to open connections and leave them in a close-wait state. After a while, the number of connections sitting in the close-wait state fills up all the connection slots that Windows has and network operations on that PC pretty much stop since nothing can create connections any more. Closing the admin console, frees all those connections and the system goes back to normal. For me, on my Windows 7 VM, it seems to take around 24 hours for it to use all the connections. It seems independent of the particular node server being used. It's pretty easy to watch it happen with a program like TCPView. It will list all the network connections and their current state and count how many connections are in each state. I wonder if that could explain the admin console just hanging randomly sometimes from the first second it opens to hours later. The logic seems right for what I have seen over the years with every version AFAICT.
bpwwer Posted June 19, 2018 Posted June 19, 2018 10 minutes ago, larryllix said: I wonder if that could explain the admin console just hanging randomly sometimes from the first second it opens to hours later. The logic seems right for what I have seen over the years with every version AFAICT. I've only seen a problem when a node server is running. I noticed it when I started developing the WeatherFlow node server and the system would have network issue after running over night. My first thought was that my node server was causing the problem but after some debugging, It turned out that running NodeLink would have the same effect. The close-wait count for the admin console process would increase whenever a node server was sending data to the ISY. Stop the node server and the close-wait count stops increasing. It's possible that other things could trigger this in the admin console too, but I haven't noticed. I've only tested this using a Windows 7 VM, so I don't know if it effects admin consoles running on other platforms. It's very easy to see the effect with TCPView https://docs.microsoft.com/en-us/sysinternals/downloads/tcpview
larryllix Posted June 19, 2018 Posted June 19, 2018 1 hour ago, bpwwer said: I've only seen a problem when a node server is running. I noticed it when I started developing the WeatherFlow node server and the system would have network issue after running over night. My first thought was that my node server was causing the problem but after some debugging, It turned out that running NodeLink would have the same effect. The close-wait count for the admin console process would increase whenever a node server was sending data to the ISY. Stop the node server and the close-wait count stops increasing. It's possible that other things could trigger this in the admin console too, but I haven't noticed. I've only tested this using a Windows 7 VM, so I don't know if it effects admin consoles running on other platforms. It's very easy to see the effect with TCPView https://docs.microsoft.com/en-us/sysinternals/downloads/tcpview IIRC there was a problem identified by P2P users that found Win 7 had limited socket connections to a low quantity, like 10? Apparently MS thought that some virus attacks would use massive socket connections to attack another user with a DoS method, swamping the other user, and the limit would slow them down to a useless level. In the end there was instructions on how to go into some setup access and change the socket limit back to a few hundred sockets. This whole idea is not very familiar to me, but apparently some apps use dozens of sockets. I assume the sockets self-terminate after some lack of usage algorithm if not closed by politely by the apps using them.
Michel Kohanim Posted June 20, 2018 Posted June 20, 2018 Hi @bpwwer, We'll look into it. With kind regards, Michel
telljcl Posted June 20, 2018 Author Posted June 20, 2018 22 hours ago, Michel Kohanim said: Hi @bpwwer, We'll look into it. With kind regards, Michel Thanks - also good to know I'm not the only one noticing this.
Michel Kohanim Posted June 21, 2018 Posted June 21, 2018 Does anyone know the URLs for these sockets? We have gone through the code and the fact that sockets are released when the Admin Console is closed just doesn't map to anything in the Admin Console. With kind regards, Michel
bpwwer Posted June 21, 2018 Posted June 21, 2018 14 hours ago, Michel Kohanim said: Does anyone know the URLs for these sockets? We have gone through the code and the fact that sockets are released when the Admin Console is closed just doesn't map to anything in the Admin Console. With kind regards, Michel Well, this is what I see in TCPView Process PID Protocol Local address Local port remote addresss remote port state --------------- ---- -------- --------------------- ---------- --------------- ------------ ---------- jp2launcher.exe 3180 TCP win7dev.bobsplace.com 60370 192.168.92.26 http CLOSE_WAIT win7dev is the machine running the admin console and the remote IP is the ISY. While I've been typing this, the count of connections in this state has increased from 60 to 130 and the only difference is the local port number. They won't go away until I close the admin console. if I stop the node server I have running the count stops increasing but none of them go away. It seems to have something to do with the connection between the admin console and the ISY but I'm not sure how to get any more information about it. Probably would need to run something like wireshark. It should be easy to reproduce unless it linked to something environmental like java version. Looks like I'm running jre 1.8.0_161
Michel Kohanim Posted June 22, 2018 Posted June 22, 2018 Hi @bpwwer, Thanks so very much. I would very much appreciate it if you could 1. Clear your error log 2. Either change the Debug Level to 3 (in the Event Viewer). Or, telnet to ISY and issue DON. The difference is that with Event Viewer, the level goes back to 0 when Admin Console is closed 3. Start the test and send the error log to support@universal-devices.com with the start/end times of when you opened/closed the Admin Console With kind regards, Michel
Recommended Posts
Archived
This topic is now archived and is closed to further replies.