Gap in recent tick data when changing instruments

[JS999]

This was an issue on NT 6.5 for a long time, and now appears to be an issue on NT 7 as well. Since you are using different historical data servers, I suspect that perhaps it might have more to do with the NT7 code (if any of it was re-used from NT 6) rather than the server.... but I am not sure about that, perhaps you re-used some of the same code from the NT 6.5 server?

When I load a new contract at times, I sometimes encounter a situation where some of the most recent tick data (in this case it is a few of the most recent 20 second bars) is missing.

When I hit "Reload Historical Data", it does not appear - the gap persists. However, it is usually moved along a bit in time. In other words, it is as though the software is creating a gap that always removes the last "X" bars of data behind the current bar.... if I was to reload the data 5 minutes later, the data that is currently missing would be filled in, and a new set of data would be missing a few bars back from wherever the market was at the time.

On NT 6.5 changing instruments and then changing back to the original instrument sometimes resolved the issue, but now nothing seems to work. This is only the case on tick-based charts. I think the bug is somewhere in your charting routines, since it is probably the case that the data is being loaded from your server, just not being displayed properly. (That would be my guess.)

You may or may not be able to replicate this error, since it only occurs periodically. You will have to keep some tick-based charts open on your systems and try switching instruments from time to time to see this occur.

Here is a screenshot of CL 05-10 so you know what I am talking about. This is a very serious problem, since displaying accurate data is the primary function of a software package, and especially on charts that are not time-based (like range or volume charts), missing this data will give an incorrect impression of where the market is, not to mention screwing up any indicators that are based on it. Please endeavor to spend some time on this issue, as this has persisted in your software for a very long time, all the way back to previous versions of NT 6.5.

[JS999]

One more thing... this problem does not occur if the data for the instrument is already in another chart or market analyzer where it is being recorded in the DB. For example, if you have a market analyzer open with 4 instruments running and switch between them, you will not notice these gaps. But if you try to switch to an instrument that is not already loaded up somewhere and recording data, then the gap problem may occur. It doesn't always happen.

[JS999]

This seems to be occurring with CL 05-10 across multiple installations of NT for me today. I am connected to Zen-Fire through Mirus futures. Also, the problem applies not only to tick charts, but minute charts as well. On a 10 minute chart it is showing incorrect open/high/low/close information for the last candle. Please investigate this today and see if you can replicate on this contract.

[NinjaTrader_Jason]

Hello JS999,

I tested this on my end multiple times using different setups and instruments, but unfortunately so far I am not able to reproduce the issue.

Unfortunately we will need to have a reproducible scenario in order to troubleshoot it. Please let us know what steps will cause the issue to occur on a consistent basis.

[JS999]

Hello JS999,

I tested this on my end multiple times using different setups and instruments, but unfortunately so far I am not able to reproduce the issue.

Unfortunately we will need to have a reproducible scenario in order to troubleshoot it. Please let us know what steps will cause the issue to occur on a consistent basis.

I think this sounds like the same problem encountered in a different thread on this forum just a couple of threads down. You suggested there that the issue may have to do with cached data. I would suggest keeping an installation running on your end and try switching instruments every once in a while, without having them already loaded in another chart or market analyzer. The problem is very likely within the code that takes the cached data and combines it with recently loaded data.

Here is the thread:

http://www.ninjatrader-support2.com/vb/showthread.php?t=26663

On another note, obviously this problem is not easily reproducible, but that doesn't mean that you should just load up your charts once, see that you can't reproduce it, and give up. It is not my job to debug your software and find a reproducible scenario for you with stuff like this, it is your job to fix it whether your customers can easily reproduce it or not.

I am sure that you are aware that some software conditions are not always easily repeatable, but that doesn't mean that there isn't a bug in your code. In fact, we are 100% certain that there IS. You can't just try this scenario once, see that you can't notice the problem, and then throw it back on the customer. This has been a problem in NT for a very, very long time, and it is very serious since it goes to the core of what your software is supposed to do (namely, display accurate data).

It is very obvious that there is a bug somewhere in your code, because I and others have been seeing it for at least a year now. I have had this problem since March 2009, and I have reported it more than once, as have other people. You (not me) need to set up a test computer and try to make this condition occur from time to time on your end, since we are 100% positive that there is a problem there, and it occurs with relative frequency.

If nothing else, try to imagine how cached data could be screwing up the loading of historical data and maybe walk over the code by hand. Maybe you'll catch something there that nobody has looked at in a long time, I don't know.

This is not a new NT 7 thing, we know it could have to do with cached data, and you've got some clues to work with... please put some actual effort into this, because this standard cookie-cutter response of "well I can't reproduce it after trying once so we're just going to leave it" is getting pretty old. This bug should have been squashed a long time ago, and probably would have been if someone had actually put some testing time into it.

[NinjaTrader_Jason]

Hello JS999,

I tried various setups, instruments and charts, I did not try it once. My colleague Josh was unable to reproduce the issue reported by Harry in the other thread after many attempts as well.

Did you ever experience the issue on a different PC?

[JS999]

Hello JS999,

I tried various setups, instruments and charts, I did not try it once. My colleague Josh was unable to reproduce the issue reported by Harry in the other thread after many attempts as well.

Did you ever experience the issue on a different PC?

Yes. Multiple PCs, multiple installations, multiple operating systems (XP, Vista, Windows 7) going all the way back to NT 6.5.1000.9 or something like that. It has not occurred often using NT 7, and doesn't occur on all instruments. Today it occurred on CL 05-10.

I think we can agree that this condition is not going to be easily reproduced. It would be nice to be able to track down the bug that way, but unfortunately it can't seem to be easily done.

Perhaps you should put some extra logging code into the area of the program that you think might be causing this problem, and run a build like that locally for yourselves. Keep it running with a cache of data for some weeks/months without clearing it out. Then if you do ever get it to occur, the trace log might show you something that could give you a clue to the problem. Also, just for a sanity check, walk over the code by hand in the area that you think could be the culprit. We have already given you some clues:

1. It has to do with cached data, probably.

2. It is in code that has not changed for a long time, so don't look at new code sections.

3. It is probably in the area of code that takes data from the DB/cache and displays it on screen, since someone else reported that the data was in fact in the DB.

4. I have told you that every time I hit "Reload Historical Data", the missing chunk of data moves forward in time (meaning that the previously missing chunk is now filled in, and a new (later) chunk of data is now missing just to the left of the currently developing price bar. (In fact, the missing data is always immediately to the left of the current price bar.) We know hitting "Reload" does not solve the problem, so start with the code that is triggered by that function and walk through it to see where it could be going wrong.

5. You know that this bug is not reproducible, so try to think of conditions that could exist within the cache loading code that would be different from time to time, if possible.

And so on... I am sure you could think of a few more things since you know the code.

Narrow things down using this method and see what you can find. Maybe something will jump out at you if you walk over the code by hand. Barring that, keep a PC running for the purposes of testing this with some extra logging code inserted, and try to reproduce it from time to time.

If a bug is not reproducible, you need to put extra error-handling code and logging code to catch and report whatever exceptions or problems may occur, because they will only occur at random intervals and will not be able to be easily repeated. Even if you try running a PC for some time and find this condition occurring at some point, it's not going to help you to find it unless you can log what the code is doing in there, line-by-line.

Non-reproducible bugs are absolutely the worst thing possible, and they are real pain to track down, but this is the only way that you are ever going to catch it. Otherwise, you will have people still reporting this a year from now.

[NinjaTrader_JoshP]

JS999,

To address your earlier concerns, when reproducing bugs we do not just test once and throw it back to the customer. Things are tested on multiple computers multiple times. When we go back to the customer it is because we were unfortunately unsuccessful and need more information. Please do not misunderstand us when we say we are unable to reproduce. We are in no way delusional in there being bugs in the software that may be hard to reproduce. Everything is noted and each case is left open when we can't nail it down. We do not just brush things off and ignore them.

This is actually why we started this beta program. We wanted to incorporate willing customers to participate in the testing of NinjaTrader 7 as beta testers. The main purpose of having beta testers is to have volunteers who are able to assist the development team iron out such bugs. As compensation, beta testers receive an early copy of the software and get a hands on preview of things to come.

With that said, what you experienced in 6.5 is probably irrelevant for NT7 because data loading in NT7 is completely different. Caching did not exist in 6.5. In working closely with Harry we have already been able to squash some variants of this, but as the nature of this area goes, it is not something that we can just throw into a catch-all bin and say its one big bug. Each case is unfortunately very sensitive to the precise settings of the customer's NinjaTrader at the moment of occurrence. We will be working closely with the testers on the other thread to see if we are able to iron out those particular cases. Hopefully the fixes on those will be beneficial to you in this situation as well. Thank you for your time and efforts.

[JS999]

JS999,

To address your earlier concerns, when reproducing bugs we do not just test once and throw it back to the customer. Things are tested on multiple computers multiple times. When we go back to the customer it is because we were unfortunately unsuccessful and need more information. Please do not misunderstand us when we say we are unable to reproduce. We are in no way delusional in there being bugs in the software that may be hard to reproduce. Everything is noted and each case is left open when we can't nail it down. We do not just brush things off and ignore them.

This is actually why we started this beta program. We wanted to incorporate willing customers to participate in the testing of NinjaTrader 7 as beta testers. The main purpose of having beta testers is to have volunteers who are able to assist the development team iron out such bugs. As compensation, beta testers receive an early copy of the software and get a hands on preview of things to come.

With that said, what you experienced in 6.5 is probably irrelevant for NT7 because data loading in NT7 is completely different. Caching did not exist in 6.5. In working closely with Harry we have already been able to squash some variants of this, but as the nature of this area goes, it is not something that we can just throw into a catch-all bin and say its one big bug. Each case is unfortunately very sensitive to the precise settings of the customer's NinjaTrader at the moment of occurrence. We will be working closely with the testers on the other thread to see if we are able to iron out those particular cases. Hopefully the fixes on those will be beneficial to you in this situation as well. Thank you for your time and efforts.

That's fine Josh... I just wanted to make sure that you guys were actually on the problem. As long as you are making serious efforts to solve it, that's all that I care about... thanks for the response.