MT5 strategy tester won't connect to cloud?

New comment
 
I can't use the strategy tester on the cloud, it randomly stopped working.

I have sufficient balance, there is no issue with connectivity, I am using a VPS and it shows I have plenty of bandwidth.

The strategy tester works fine on local agents but it is really slow so I want to use the cloud network.

I don't get any errors, it starts synchronising the history and then doesn't do anything else.

 
Jake Freeman:
I can't use the strategy tester on the cloud, it randomly stopped working.

I have sufficient balance, there is no issue with connectivity, I am using a VPS and it shows I have plenty of bandwidth.

The strategy tester works fine on local agents but it is really slow so I want to use the cloud network.

I don't get any errors, it starts synchronising the history and then doesn't do anything else.

Hi Jake,

If I understand the issue correctly, I think it may be related to how many optimizations you are trying to make, during the test; I have had the same issue if my optimizations of a test exceed 100 million. 

The guide to using the MQL5 cloud service says, (at https://www.metatrader5.com/en/terminal/help/mql5cloud/mql5cloud_use),

"Features of the MQL5 Cloud Network

The entire power of the MQL5 Cloud Network is used only for complete slow optimization.

During genetic optimization, only agents of one access point are used. It is connected with the specific features of the genetic algorithm.

The genetic optimization mode is automatically enabled when the total number of optimization steps exceeds 100 million."

So, if tyou are above 100 million, it automatically switches to genetic algo testing, instead of complete slow optimization. 

In that case, it will not use the the full list of agents, just the closest group, and only some of them.

For example in my testing (when above 100 million optimizations) I typically only get about 400 out of the 14,000 available agents, at my closest group of agents; sometimes much less than that.

So, if you want to use more groups of agents (all of Europe and the USA), you will need to reduce your optimizations to below 100 million.

An easy way to do that is, for example, to test SL and TP in 5 or even 10 pip increments, instead of 1 pip increments. This alone will cut your optimizations a lot. 

The same is true for other values that you are testing for; make your increments bigger, and you can get the total optimizations below 100 million, and then be able to use the biggest number of agents.

I hope this helps!

How to Use - MQL5 Cloud Network - MetaTrader 5 Help
  • www.metatrader5.com
The MQL5 Cloud Network allows you to quickly optimize your Expert Advisors using the power of thousands of computers. The network combines remote...
 
flya200 #:

Hi Jake,

If I understand the issue correctly, I think it may be related to how many optimizations you are trying to make, during the test; I have had the same issue if my optimizations of a test exceed 100 million. 

The guide to using the MQL5 cloud service says, (at https://www.metatrader5.com/en/terminal/help/mql5cloud/mql5cloud_use),

"Features of the MQL5 Cloud Network

The entire power of the MQL5 Cloud Network is used only for complete slow optimization.

During genetic optimization, only agents of one access point are used. It is connected with the specific features of the genetic algorithm.

The genetic optimization mode is automatically enabled when the total number of optimization steps exceeds 100 million."

So, if tyou are above 100 million, it automatically switches to genetic algo testing, instead of complete slow optimization. 

In that case, it will not use the the full list of agents, just the closest group, and only some of them.

For example in my testing (when above 100 million optimizations) I typically only get about 400 out of the 14,000 available agents, at my closest group of agents; sometimes much less than that.

So, if you want to use more groups of agents (all of Europe and the USA), you will need to reduce your optimizations to below 100 million.

An easy way to do that is, for example, to test SL and TP in 5 or even 10 pip increments, instead of 1 pip increments. This alone will cut your optimizations a lot. 

The same is true for other values that you are testing for; make your increments bigger, and you can get the total optimizations below 100 million, and then be able to use the biggest number of agents.

I hope this helps!


I don't think it has to do with the amount of operations because it won't even do a backtest or allow me to use the market scanner.
 
Jake Freeman #: I don't think it has to do with the amount of operations because it won't even do a backtest or allow me to use the market scanner.

Single back-tests don't use the MQL5 Cloud. Only optimisations use the MQL5 Cloud if the number of passes is greater that what the local computer can handle.

 
Fernando Carreiro #:

Single back-tests don't use the MQL5 Cloud. Only optimisations use the MQL5 Cloud if the number of passes is greater that what the local computer can handle.

The same thing happens with market scanner and optimisations, I can't get any operation to use the cloud as it would normally
 
Jake Freeman #: The same thing happens with market scanner and optimisations, I can't get any operation to use the cloud as it would normally

There is insufficient information to provide a suitable answer. Are there any messages in the Strategy Tester log, or the MetaTrader log?

Showing those logs might help!

EDIT: Also, given that it is a VPS, there may be firewall issues too. Analyse the firewall logs to see if the communications are being blocked.

 
Fernando Carreiro #:

There is insufficient information to provide a suitable answer. Are there any messages in the Strategy Tester log, or the MetaTrader log?

Showing those logs might help!

EDIT: Also, given that it is a VPS, there may be firewall issues too. Analyse the firewall logs to see if the communications are being blocked.

I've taken a screenshot of what happens when I run the market scanner. After this point it just sits doing nothing. That's what's confusing me the most, there is no error, it just sits doing nothing.

I will look at the firewall logs too but I don't think anything will be blocked in there. I use ForexVPS, it wouldn't make any sense for them to block MT5 services.

Files:
Journal.jpg  523 kb
 
Jake Freeman #:
I've taken a screenshot of what happens when I run the market scanner. After this point it just sits doing nothing. That's what's confusing me the most, there is no error, it just sits doing nothing.

I will look at the firewall logs too but I don't think anything will be blocked in there. I use ForexVPS, it wouldn't make any sense for them to block MT5 services.

What does the MetaTrader Journal log report (not the Strategy Tester Journal log).

What is the amount of available disk space at that time?

What is the CPU and RAM usage at that time?

 
Fernando Carreiro #:

What does the MetaTrader Journal log report (not the Strategy Tester Journal log).

What is the amount of available disk space at that time?

What is the CPU and RAM usage at that time?


I checked the MT5 journal, it doesn't update at all when I try to run the test.

The disk space is fine. (80gb of space)

Task manager shows totally stable.

I also disabled the firewall and nothing has changed.


Obviously I have also tried restarting the VPS.
 
Jake Freeman #: I checked the MT5 journal, it doesn't update at all when I try to run the test. The disk space is fine. (80gb of space). Task manager shows totally stable. I also disabled the firewall and nothing has changed. Obviously I have also tried restarting the VPS.

Try doing a simple optimisation on a single symbol, to see if it works. The optimisation on multiple symbols may be overloading something causing it to slow down too much.

If the simple optimisations works then redo the large one but wait a very long time to see if it responds.

One more point I would like to add, is that if the VPS is using a SSD, then the pause may be because of a TRIM sequence on the drive. With so many large files being generated and deleted, it may initiate a TRIM sequence on the drive causing the "pause". You can detect this by looking at the Task Manager on the Performance graph of the disk drive and if it's usage is at 100% but the actual Read and Write speed are very low, then a TRIM sequence is taking place on the drive.

 
Fernando Carreiro #:

Try doing a simple optimisation on a single symbol, to see if it works. The optimisation on multiple symbols may be overloading something causing it to slow down too much.

If the simple optimisations works then redo the large one but wait a very long time to see if it responds.

One more point I would like to add, is that if the VPS is using a SSD, then the pause may be because of a TRIM sequence on the drive. With so many large files being generated and deleted, it may initiate a TRIM sequence on the drive causing the "pause". You can detect this by looking at the Task Manager on the Performance graph of the disk drive and if it's usage is at 100% but the actual Read and Write speed are very low, then a TRIM sequence is taking place on the drive.


I tried running a simple optimisation, this time the cloud says processing but doesn't do anything, so there is a change but it doesn't do anything.

The optimisation is only 130 passes so should be over in no time. It freezes in exactly the same way but it does acknowledge the optimisation on the cloud and begins processing.

The disk usage is at a totally stable number.
12
New comment