Hey folks, there's been a parallel thread at: http://finance.groups.yahoo.com/group/TWSAPI/message/16968 Today I found that I can reproduce the problem at will. Here's the procedure: 1) use Sysinternals TCPView to close the connection made by TWS to gw1.ibllc.com:4000. This generates the "connectivity lost" message. 2) After TWS reconnects I then place an order. 3) The order is filled, TWS shows an updated position, but no trade. My program doesn't receive the fill. 4) I restart TWS, and now it shows the fill. I did this twice, once at 8:22 PST to buy 100 ZION, the next at 8:51 PST to sell 100 ZION. Both times the result was identical. I've communicated this to IB. Hopefully we'll get a quick response. Regards, -Russ
Nice work, Russ! It seems like that ought to be helpful to the IB programmers. Aaron Schindler Schindler Trading
It appears that TWS version 893 has fixed the problem, according to after-hours testing in a paper trading account. We'll see for sure on Monday. Regards, -Russ
I still had the problem after first upgrading to Java 1.6.0_13 and then upgrading to TWS 893.5 (March 25). So if you don't still have the problem tomorrow (Monday) I would say it is a TWS server thing that has been fixed and nothing to do with version 893.5. I haven't risked leaving TWS running from the nightly reset to the start of trading since having this problem three days in a row in late March. On or around March 22nd IB changed something such that TWS needs to be restarted sometime between the nightly reset and any API trading, or else TWS won't show executions and correct positions. Aaron Schindler Schindler Trading
we are still experiencing missing data. trade executions are still not showing after entry or exit and the trade stays frozen on the screen until you log off and log in. It has taken several attempts to log in and or to remove frozen order. This is extremely frustrating when trading through the night for active traders. One trade you can remember, 100 trades buying and selling you are not sure if you are in or out, part filled or your what your true average price is. especially dangerous as the traders are trading the futures through the night. IB as a discount broker the costs of doing business are starting to make you NOT A DISCOUNT. W
A lot of people here on ET as, well as on other forums, are stating that IB notified them that TWS XXX would no longer be supported and they need to upgrade. This means that the server end at IB was going to change to eliminate such support since the users already had a good version of TWS. IB almost never admits there are problems in their software, they just issue another update. I suspect that the rash of problems that are being reported are more the result of server changes than are due to TWS 893.x. We are just acclimated to the TWS version change horrors. Once we connect to IB we are probably internally routed, based internal loading, to servers with different versions of the server software (IB may deliberately update servers incrementally or just because someone messes up). Thus, a TWS restart may cause a reconnect to a different internal server and what appears to be a cure. Of course a reconnect the next day may be back to a bad server. Jack
Well, it appears to be fixed anyway. Version 893.5 (Apr 2) is still working well on Monday morning, after starting it Sunday night and leaving it running through the midnight disconnect. I also tried disconnecting it using Sysinternals TcpView, and it reconnected properly, showing execDetails messages for subsequent orders. For kicks, I also tried version 891 again, and it worked just fine after a disconnect. Go figure. So apparently IB made some backend changes that broke the reconnect functionality. Now they've fixed it, at least for me. Let's hope that nothing else breaks in the future. It's a complete PITA. Cheers, -Russ
IB says that the problem appears to be caused on their backend by some advanced features. Currently they have turned off these features while they are debugging the problem. That probably explains why systems are working properly now. They aren't going to turn the features back on until the problem is fixed. Regards, -Russ
I'm glad I put the almost in there. Now that the've acknowledged the problem I'm willing to try Java 1.6.0_13. Jack