Hello BloodyRookie so here is the first test of my supernode https://supernodes.nem.io/details/843
I see that my supernode failed. I checked all tabs and it seems that the only one condition my supernode did not pass was responsiveness. Am I right?
How to improve responsiveness to pass supernode conditions? Is it about computonial power or is is about ISP?
As i wrote above, wait 24 hours, it is normal that the node fails in the first 2 rounds.
1 Like
OK, I will be waiting.
Hi BloodyRookie, I know that you told that it is normal that supernode fails in first two rounds but my supernode https://supernodes.nem.io/details/843 have just failed also on third round at the same criterium RESPONSIVENESS. Do you still think it is OK (normal) and it will be OK? I have some bad feeling that there is some problem but I do not know where.
Thanks for your opinion and I hope you can solve it.
hmm…yea, that does not look good. How much RAM do you give to NIS?
I did not change anything about RAM management in config.propereties (default settings). Or is there some settings I can try to change to solve it?
BTW, my node is runnig on Sony Vaio S15, with 512GB SSD and 12GB RAM. From this point of view I cannot see any performace issues to running supernode.
If you have 12GB RAM available, you should change the NIS start script to use more RAM, like 8GB and you probably should use G1 garbage collector. Something like adding
-Xms8G -Xmx8G -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:+PrintGC -Xloggc:"./gc.txt"
in the start script which also logs the some garbage collector information to see how well it is doing.
I do not have 12GB free RAM! 12GB is my total RAM for my notebook. So this will eat most of my RAM :). So I hope that 2G should be enough.
So this is my actual configuration of NIS.bat file:
pushd nis
java -Xms512M -Xmx1G -cp “.;./;…/libs/” org.nem.deploy.CommonStarter
popd
Could you please tell me what the number after Xms means and what the number after Xmx means? I want to better understand this values to properly set up right values for my computer.
I have also another idea. On my Asus router I have turned on DoS protection. So I tried to turn it off and I will see if that can help. What do you think?
This is what DoS protection on Asus router means https://www.asus.com/support/FAQ/1031610/
It seems that turned off Asus DoS setting did not help. I turned it off before round 2622 and as you can see it did not help.
So, now I am trying to set and run NIS with -Xms1G -Xmx2G. I hope that two times higher setting compared to default settings will help. If not what will be your opinions?
Is not this a matter of network quality?
It is highly likely that packet loss is occurring somewhere.
I really do not know. That is why I am asking here. So do you think I should try to solve this issue with my ISP? And ahat should I tell to my ISP? That I have a packet loss or what? And it is not a problem of any settings of my NIS? Thanks.
I do not know if ISP can get the answer.
From the test result of SuperNode, the following possibilities are considered.
There is a problem somewhere from the network interface of your PC to your ISP.
Moreover, it may be also in ISP.
PC main unit · Network cable · Router · Connection line to ISP.
Do not have any problem with any one.
As a separate case, I think that the result will be better if the starting options are aligned with Xms and Xmx.
-Xms2G -Xmx2G
Also, if necessary, it is good to add -XX:+UseG1GC
Here, improvement of response performance is expected.
thanks
Thanks for your opinions. But as you can see I have too much choices :). But still I hope that I will solve it with your (or BloodyRookie) help.
First of all I will try -Xms2G -Xmx2G setting.
I want to ask you what this -XX:+UseG1GC exactly means and what is the actual impact of this setting on performance of my node?
In Java, JVM is managing memory.
It is necessary to arrange and release the memory which became obsolete and no longer needed.
For that option, we changed the algorithm for releasing memory to G1GC.
In Java 8, it normally moves at high speed, and the program stops for a while during memory release. That is the default setting.
So, using G1GC, the program will stop for a minimum amount of time.
The program will not stop for more than 200ms (G1GC default) as a standard value.
Currently there is almost no single CPU & single threaded environment, so there is almost no stopping.
Basically, while executing the program, we also proceed with releasing unnecessary memory at the same time.
However, since the effect becomes bigger when consuming more memory, there is no guarantee that the effect will be comprehensively effected in this case.
In Java 9, G1GC becomes the memory release algorithm at normal time.
thanks
Thanks for clarification mizunashi.
For round 2623 I had -Xms2G -Xmx2G setting. But as you can see it did not help.
So now for next round 2624 I made this settings (from BloodyRookie opinion):
java -Xms2G -Xmx2G -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:+PrintGC -Xloggc:"./gc.txt" -cp “.;./;…/libs/” org.nem.deploy.CommonStarter
So I can post here also gc.txt file after rund 2624.
But I have a bad feeling that this will not help to solve my responsiveness problem.
Interesting is that for round 2622 I had the best responsiveness time 1333 ms (6/10) and it was with changed setting of DoS in my Asus router (I turned off DoS protection for that round).
So now I again turned off DoS setting in my Asus router to check if it helps and also I turned off QoS setting. Maybe this should has also impact to pass responsiveness test.
But if this do not help I really do not know what to do next.
BloodyRookie, do not you think that problem should be that my node (I am located in Czech Republic - Europe) is testing by node that is too far away from my node? E.g. from USA or something like that? Could you please check it if this is not the problem?
Doesn’t seem to be a memory problem. The ping test also shows problems: -1; -1; 88; 48; 42
Meaning that 2 out of 5 times the servant did not respond to the ping in a timely fashion.
what is the node’s public ip, 213.192.24.194 or 213.192.4.194? Both IPs are from Czech Republic but only the latter can be pinged regularly. You mentioned that there first was a provider problem, what exactly was it?
Node’s public ip is 213.192.24.194. Not 213.192.4.194, where you can see that IP at the moment?
When provider gave me public IP address (213.192.24.194) he made soem typo in (I do not know exactly) som table and when I tried to check what is my IP I still saw the old IP address which was not public (and that is the IP 213.192.4.194). But after that I called my ISP to tell him this and then I repair it and told me that he made only some typo.
So now you should see only one IP (public) for my node, which is 213.192.24.194. But you wrote that this IP has some ping issues which should be linked to responsiveness issues my node has, right?
So the last round 2624was totally bad. And this is what I get with best setting (at least I think so). With 2GB RAM and also with added G1GC option.
So I really do not know what is wrong now. I am going to set my node and router as it was before (active DoS protection, and QoS) and also I try to contact my ISP. But it is really weird problem I think.
Hi, so another day is over, I have some communication with my ISP and he do not know what is wrong. Everything should be OK from his side. And I really donot know what should I try to tweak. Could you please cheeck it more detaily if you can BloodyRookie? Because I cannot see the reason why it is not working properly. Thank you very much.
https://supernodes.nem.io/details/843
Looks like your node doesn’t like to get a bunch of requests at the same time. But i have no idea what is causing the problem. It is as if your node was behind the great chinese firewall. Those guys had those kind of problems too.