AutoRestart token-based re-login fails on unfunded paper accounts

[neyoneit]

IBC version: 3.23.0
Gateway version: 10.37.1o (build 1037)
OS: Windows Server 2025 Datacenter 10.0.26100

Setup:
I run 3 separate IB Gateway instances on 3 separate IBKR accounts (each with its own credentials and settings directory). All accounts are unfunded with paper trading enabled. Each instance has its own IBC config, port, and AutoRestartTime staggered 5 minutes apart.

Problem:
Full login with credentials works fine every time. However, when AutoRestart triggers and uses the token file, IBC clicks "Re-login" without filling in username/password, resulting in "Unrecognized Username or Password". The Gateway then gets stuck on this error dialog permanently.

Reproduction (happens consistently on 2 out of 3 instances):

Start Gateway with IBC — full login with credentials succeeds
AutoRestartTime triggers, Gateway shuts down
IBC finds autorestart token file (autorestart file found)
Gateway starts, shows "Re-login is required" dialog
IBC clicks "Re-login", but does NOT fill username/password
"Unrecognized Username or Password" error, Gateway stuck forever
From the attached log (port 4003):
IBC-3.23.0_GATEWAY-1037_.txt

23:05 — Fresh start: Setting user name → Setting password → Login has completed ✅
23:50 — AutoRestart: autorestart file found → Re-login is required → clicks Re-login → Unrecognized Username or Password ❌
I have logs from two instances (ports 4003 and 4005, different IBKR accounts) showing the exact same pattern.

I believe AutoRestart should work on paper accounts the same way it does on funded accounts. Paper trading is essential for testing strategies before going live.

[rlktradewright]

Ok, several points to make here.

First, autorrestart works fine under IBC for paper trading accounts linked to funded accounts. I run my paper trading account under IBC all day every trading day with no problems.

Next, until a few days ago, I wasn't aware that IBKR would let you have an unfunded account with paper trading enabled, but it may just be that no-one has used IBC with such an account until very recently.

What has been available for years is a free trial account, which is just a paper trading account (not linked to a 'real' account) . With these accounts, auto-restart just doesn't work. But it's not a fault in IBC: rather the autorestart functionality in Gateway/TWS simply doesn't work, and is presumably not intended to work.

I suspect the same is true for unfunded accounts. If that's the case, then I can't make IBC do auto-restart for these accounts.

So you could help here by running Gateway without IBC but with autorestart time set in the Lock and Exit settings, and running it to see whether autorestart works properly in the absence of IBC:

• If it doesn't, work, then you're out of luck,
• If it does work, then I'll have to see what I can do to make it work under IBC. You could also help here by investigating what happens if you enter the username and password in the 'Unrecognized Username or Password' dialog: does it all continue nicely, or what?

Note however that if I can't make it work with IBC, it's not really that much of a problem: all you need to do is set the auto-logoff time, and user Task Scheduler (or cron on Linux) to restart it. Your API apps will need to be able to cope with the loss of connection to Gateway/TWS during the restart, but they probably need to be able to do this anyway.

Where auto-restart is most valuable is when using 2FA, which is mandatory for live accounts but is not used at all by paper-trading accounts.

[neyoneit]

Thanks for the response. I've implemented a workaround - a PowerShell watchdog that monitors IBC logs for "Unrecognized Username or Password" and automatically kills + restarts the Gateway when detected. This works reliably.

I haven't (cant remember if i tried honestly now) tested manually entering credentials into the error dialog yet - I'll try that next time it happens and report back (ill disable my workaround to test it).

What's puzzling is the inconsistency: I run 3 separate Gateway instances on 3 separate unfunded paper accounts (all configured identically). When AutoRestart triggers, sometimes 2 out of 3 recover fine and one gets stuck on "Unrecognized Username or Password". It seems almost random which instance fails, like a lottery.

Same thing with the Login Messages dialog I reported in #343 - it only appears on some instances, some of the time, despite identical configuration. Very hard to reproduce reliably.

These might just be quirks specific to unfunded paper accounts. I'll keep testing and update this issue if I learn more.

[neyoneit]

I can't get this to work reliably, honestly, every workaround I try falls apart eventually. I'd really appreciate it if you could create a paper account on new unfunded account and test it, because I can't be the only one using multiple unfunded paper accounts to test different strategies, not just one, but several.

[rlktradewright]

Several things to say here:

1. You don't need multiple accounts to run several strategies. You can run all your strategies in one account. You just need to keep track of which strategy created each order.
2. I don't want to waste my time investigating unfunded accounts. If you can't afford to put a minimal deposit into the account, then you can't afford to trade. Don't be a cheapskate. If something doesn't work with a funded account, I'll be happy to help.
3. I've made some suggestions which you seem to have ignored.

• "you could help here by running Gateway without IBC but with autorestart time set in the Lock and Exit settings, and running it to see whether autorestart works properly in the absence of IBC"
• "you could also help here by investigating what happens if you enter the username and password in the 'Unrecognized Username or Password' dialog: does it all continue nicely, or what?"
• "all you need to do is set the auto-logoff time, and use Task Scheduler (or cron on Linux) to restart it. Your API apps will need to be able to cope with the loss of connection to Gateway/TWS during the restart, but they probably need to be able to do this anyway"

Regarding that final bullet point: before IBKR introduced the auto-restart functionality, that was how I ran both my live and paper accounts. Although there wasn't any mandatory requirement for a daily restart of TWS/Gateway back then, I found that things were more stable running like that rather than just leaving TWS up all day 5 days a week. Autorestart provided a better solution, but if it doesn't work with your account type thaen this is the easiest way to go.

[neyoneit]

The "cheapskate" comment is outright disrespectful and frankly embarrassing coming from a maintainer of a public open source project. You should know better.
Let me be clear about what happened here: you admitted you didn't even know how to reproduce these issues or if they existed until I showed up. I did all the work, reproduced the issue across multiple instances, provided detailed timestamped logs, documented the exact failure pattern, and handed you a complete, actionable bug report. You contributed nothing to the investigation and responded by questioning my finances. That's not how you treat contributors.
And the "paper trading is only for accounts linked to funded accounts" argument is nonsense. Paper trading exists specifically so people can test strategies before risking real money. That is its entire purpose. IBKR offers it, your README says nothing to the contrary. If you've silently decided unfunded paper accounts are second class citizens not worth your time, that is a decision you made without telling anyone, and people are wasting hours because of it.
Document it. Clearly. In the README. Which account types work, which don't, and why. If the honest answer is "because I can't be bothered", write that too. At least people will know what they're walking into instead of spending their time filing bug reports you're just going to sneer at.

[rlktradewright]

"The "cheapskate" comment is outright disrespectful and frankly embarrassing coming from a maintainer of a public open source project. You should know better."

• If you're not a cheapskate, then stick $100 in your account and see if that solves the problem, instead of trying to get me to spend my valuable unpaid time solving a problem that I don't have and cannot reproduce. I'm not embarrased by the comment, disrespect is in the eye of the beholder, and perhaps you need to review your attitude of entitlement to those who work for nothing on your behalf.

"Let me be clear about what happened here: you admitted you didn't even know how to reproduce these issues or if they existed until I showed up"

• That's true, I didn't know what the cause of these reports was (and you're by no means the first person to raise this issue) until you, very helpfully, stated that you were using unfunded accounts. That was the first time anyone had made a suggestion that might throw a light on this. Given that information, I suggested you should run Gateway/TWS without IBC but with autorestarttime set, to see if autorestart still failed (just to make sure it wasn't a fault in IBC causing this). But you don't appear to have folllowed this simple suggestion, so I still don't actually know whether the unfunded aspect of your account is relevant.

"I did all the work,"

• Apart from the many hours I've spent over the past year or two puzzling over this issue from previous reports by other users. Don't underestimate the time it consumes trying to anylyse logfiles for any clue as to what might be going on.

"provided detailed timestamped logs"

• Er, excuse me, it's IBC that creates the detailed timestamped logfiles: adding them to your Github post is hardly an effort.

"handed you a complete, actionable bug report"

• This is not a bug in IBC. Your report didn't add anything new to what had been reported before, except the (possibly) crucial fact that you're using an unfunded account. There was nothing actionable about it - if there had been something new I could have done as a result of your report, I would have done it. This whole repository is a testament to the fact that I take issues seriously and take action where needed and possible. It's not possible to take action where I don't have any evidence to go on, especially when the cause appears to be something within IBKR's system.

"_ You contributed nothing to the investigation_"

• Now you're just being insulting.

"responded by questioning my finances"

• No I didn't. I know nothing about your finances. But it's certainly true that if you can't fund an account, then you can't afford to trade. That's just a truism.

"That's not how you treat contributors."

• Thank you for your advice.

"the "paper trading is only for accounts linked to funded accounts" argument is nonsense"

• I never suggested any such thing. Here is a simple fact: if you have a funded account, you get a paper-trading account linked to it., and autorestart in such accounts works perfectly, whether or not IBC is in use . I run my live and paper accounts all day every trading day under IBC with autorestart, wihtout any problems. And here's another statement that is still a conjecture: paper trading accounts linked with unfunded main accounts don't appear to support autorestart. I'm asking you to confirm or deny this hypothesis by running your account without IBC, as mentioned before, and if that fails, then to fund the account to a fairly minimal level, and see if that corrects the problem.

"If you've silently decided unfunded paper accounts are second class citizens not worth your time, that is a decision you made without telling anyone, and people are wasting hours because of it."

• I think you're losing the plot here . If IBKR decide not to support autorestart for paper trading accounts linked to unfunded main accounts, that is nothing to do with me. If it's true (and as I said above, it's still a conjecture), then of course I can document it appropriately. And once I know it's true, I can consider what action IBC could take to ease the situation for users of such accounts (which takes me back to one of the other things I suggested you could do - check whether filling in the username and password after receiviing the 'Unrecognized Username or Password' dialog results in proper login, because that is something that IBC could certainly automate).

"Document it. Clearly. In the README. Which account types work, which don't, and why. If the honest answer is "because I can't be bothered", write that too. At least people will know what they're walking into instead of spending their time filing bug reports you're just going to sneer at."

I disagree. This is for IBKR to doocument, not me. As for 'sneering', I'll let that pass as your understanding of the langauge seems to be different from mine..

[neyoneit]

Let's go point by point, since you seem to enjoy that format.

"Stick $100 in your account." You still don't get it. This is not about whether I can afford to trade. Paper accounts exist precisely so that people can evaluate strategies before putting any real money on the line. That is their purpose. Suggesting I deposit money to test a paper trading bug is like telling someone to buy a car to test a car simulator.

"It's IBC that creates the timestamped logs, adding them is hardly an effort." Right. And yet previous reporters of this exact same issue apparently didn't do it, which is why you spent "a year or two puzzling over logfiles" and still had no idea what was going on. I show up, hand you the logs, identify the account type, narrow down the failure pattern across three instances, and you respond by telling me that attaching a file to a GitHub issue isn't impressive. If it's such an easy thing to do, why didn't you do it already?

You also mention I'm "by no means the first" to report this. Correct. Multiple people reported it, you spent a year or two on it, and still had nothing. I show up, identify the account type, reproduce it across three instances in a week, and apparently add nothing new. Remarkable.

"I'm not the only one." And neither am I. While you've been busy lecturing me about funding accounts, someone else hit the exact same problem, diagnosed it, fixed it, and submitted PR #344 a week ago. It's sitting there unreviewed. So the community is actively solving this for you and you still can't find the time to look at it, but you did find the time to write multiple paragraphs telling me to deposit money. That says everything about where your priorities are.

"This is not a bug in IBC." Maybe. But IBC is the layer between the user and Gateway, and IBC is what clicks Re-login without filling in credentials. Whether the root cause lives in IBKR's system or yours is exactly what investigation is for, which requires someone to actually look at it. PR #344 suggests it very much can be addressed on the IBC side.

You used the word "entitlement". Let's talk about entitlement. Expecting contributors to do your debugging, write your bug reports, and now apparently also fix the code and submit PRs, while you lecture them about their bank balance and let those PRs rot unreviewed because the author happened to use an unfunded paper account, that's a peculiar definition of working "for nothing on their behalf".

"Documentation is for IBKR, not me." You maintain a tool. Users of your tool run into this. You know it happens. Writing one sentence in the README costs you nothing and would save the next person this entire conversation with a fool who can't give a damn about approving a pull request that apparently already solves the problem. And that argument doesn't hold up anyway, because IBC is not IBKR's product. You decide what it supports, you know what doesn't work, and it's your README that says nothing about it. Passing the buck to IBKR for documentation of your own tool's limitations is convenient, but it's not an argument.

And as for my "understanding of the language", coming from someone who in the same breath wrote "doocument" and "anylyse" and "thaen" and "wihtout", I'll take that one on the chin.

[rlktradewright]

Yes, point by point seems a very appropriate format in this sort of situation. It isn't always.

"Stick $100 in your account." You still don't get it. This is not about whether I can afford to trade. Paper accounts exist precisely so that people can evaluate strategies before putting any real money on the line. That is their purpose. Suggesting I deposit money to test a paper trading bug is like telling someone to buy a car to test a car simulator.

• Yes, well, I've been using the paper trading system since it waa first introduced in, I think, 2006, so I think I have a reasonable understanding of what it's for. And as I indicated in a prrevious post somewhere, it's only very recently that I became aware that IBKR offer a paper trading account with an unfunded main account, but that may well have been just an oversight on my part.
• What this is about is the conditions under which a paper trading account works exacty like a 'live' account. Your analogy about the car simulator is flawed. A better analogy would be the car salesman omitting to tell you that if you want to be able to run the car on more than one day, you have to pay a refundable deposit of, say, $200.
• I would imagine that maintainng the paper-trading infrastructure is a considerable cost to IBKR, so I don't think it's unreasonable for them to apply a minor restriction to paper trading accounts linked to non-funded main accounts. Though I suspect it's more likely that this is a simple bug, so why don't you report it to them (once you've done the decent thing and tried running that paper trading account without IBC and using autorestart)?
• By the way, the paper trading account implementation is much more realistic than a car simulator. In a car simulator, nothing is real (except perhaps the pedals and other controls). In the paper trading system, it's all the same code as the live system, except that order matching is done in the IBKR servers rather than at a real exchange, and the account doesn't contain real money.

"It's IBC that creates the timestamped logs, adding them is hardly an effort." Right. And yet previous reporters of this exact same issue apparently didn't do it, which is why you spent "a year or two puzzling over logfiles" and still had no idea what was going on. I show up, hand you the logs, identify the account type, narrow down the failure pattern across three instances, and you respond by telling me that attaching a file to a GitHub issue isn't impressive. If it's such an easy thing to do, why didn't you do it already?

• Previous reporters of this issue did send me log files. As I said, I've spent many hours poring over them. The reason I had no idea what was going on was because the only discernible difference between an autorestart log for a live account or a paper account linked to a funded live account, and an autorestart log for a paper account linked to an unfunded live account, is the appearance of the 'Relogin is required' dialog at some point after autorestart has apperently succeeded in the lunfunded case.
• Saying 'If it's such an easy thing to do, why didn't you do it already?' is a complete non-sequitur. I have no need to siubmit logs about this issue because I haven't experienced the issue, so what;'s your point.

You also mention I'm "by no means the first" to report this. Correct. Multiple people reported it, you spent a year or two on it, and still had nothing. I show up, identify the account type, reproduce it across three instances in a week, and apparently add nothing new. Remarkable.

• Yes, you showed up and reported the same problem as others but with the interesting fact that you were using an unfunded accountt, that no-one else had mentioned. Does that make you some kind of a genius or something? If you hadn't mentioned that, then we wouldn't have moved forward at all, because that was the only thing in your report that was different.

"I'm not the only one." And neither am I. While you've been busy lecturing me about funding accounts, someone else hit the exact same problem, diagnosed it, fixed it, and submitted PR #344 a week ago. It's sitting there unreviewed. So the community is actively solving this for you and you still can't find the time to look at it, but you did find the time to write multiple paragraphs telling me to deposit money. That says everything about where your priorities are.

• PR344 is not sitting there unreviewed. Look again.
• Someone didn't diagnose the problem or fix it. They applied a brute force AI-generated kludge to provide a workaround.. Yes I know it deals with the immediate issue by merely forcing a cold restart, but as my response in the PR makes clear it's inadequate as a 'production-ready' fix.
• Thank you for acknowledging the fact that writing these replies takes time. It certainly does!
• And thanks for your armchair psychoanalysis of me, but really, you're not very good at it...

"This is not a bug in IBC." Maybe. But IBC is the layer between the user and Gateway, and IBC is what clicks Re-login without filling in credentials. Whether the root cause lives in IBKR's system or yours is exactly what investigation is for, which requires someone to actually look at it. PR #344 suggests it very much can be addressed on the IBC side.

• IBC clicks Re-Login 'without filling in credentials' because it's the only thing it can do. The 'ReLogin is required' dialog coontains no credentials fields, no other controls at all.
• If you take the trouble to look back through the whole history of this issue, both here and in the IBC User Group, you'll see that several times I've suggested a possible resolution might be to fill in the credentials in the 'Unrecognized Username or Password" dialog that eventually apperars, to try and re-initiate login. But I'm not going to do that unless I know it has a chance of success. But to assess that chance, I need to get someone who experiences the issue to actually do this and report what happens, as I've asked you to do. But for some reason, no-one (inclding you) seems to want to do this. I don't know why: it would take two minutes of their time and would provide valueable feedback

You used the word "entitlement". Let's talk about entitlement. Expecting contributors to do your debugging, write your bug reports, and now apparently also fix the code and submit PRs, while you lecture them about their bank balance and let those PRs rot unreviewed because the author happened to use an unfunded paper account, that's a peculiar definition of working "for nothing on their behalf".

• This is such utter nonsense that I'm not even going to bother (because obviously I don't give a damn...).

"Documentation is for IBKR, not me." You maintain a tool. Users of your tool run into this. You know it happens. Writing one sentence in the README costs you nothing and would save the next person this entire conversation with a fool who can't give a damn about approving a pull request that apparently already solves the problem. And that argument doesn't hold up anyway, because IBC is not IBKR's product. You decide what it supports, you know what doesn't work, and it's your README that says nothing about it. Passing the buck to IBKR for documentation of your own tool's limitations is convenient, but it's not an argument.

• Ok, what do you think I should say in the README? How about this:

A tiny percentage of users experience a problem where, some time (usually a small number of hours) after auto-resatart, TWS/Gateway display a dialog entitled "Unrecognized Username or Password" and the program sits waiting for user action. It is not known what causes this issue but available evidence indicates that it is due to some condition that the IBKR login server takes note of but doesn't explain."

Since hardly anyone experiences this (at least until very recently, and it's still only a handful), so most users will never see it, and since it makes no suggestionas as to what to do to avoid it or work around it, I don't think that would be helpful.

And as for my "understanding of the language", coming from someone who in the same breath wrote "doocument" and "anylyse" and "thaen" and "wihtout", I'll take that one on the chin.

• Yes, these are called typos. They are errors that I make just as much as anyone else (possibly more so as my eyesight is not all it used to be (actually it never was very good)!. They are not slurs like 'sneering'.

I don't see m,ch point in us continuing with this.

One final point, you took me to task for using the work 'cheapskate'. Actually I only said 'don't be a cheapskate'. On the other hand, you have been consistenly rude to me all the way through. You should know better.

[neyoneit]

"AI-generated kludge." Right, because a working fix that solves the problem is a kludge the moment it mentions Claude in the commit message. It's telling that out of everything in that PR, that's what stood out to you. Clearly it isn't irrelevant to you, which says more about your insecurities than about the quality of the fix.
"Hardly anyone experiences this." In the span of one week, two separate people showed up independently with the exact same issue, one of them submitting a PR. How many more quietly gave up, worked around it, or never bothered reporting it? "Hardly anyone" is not a data point, it's cope.
Your README suggestion is pathetic. "A tiny percentage of users experience" after two years, two reporters in one week, and a working PR sitting in your queue, you write that nonsense. And speaking of that PR, you asked @meligo what type of account they use, sheesh i wonder why - maybe to suggest he fund the account instead of resolving the issue, right. Same as #345. "let me find a reason to close it without engaging." "Everything has a solution." Yours is always the same one: throw money at it and the problem disappears. That's not a solution, that's a laziness.
And the entitlement comment. The word you're looking for is mirror.

Feel free to close this issue same way you're gonna treat the pr.

[rlktradewright]

Ok, I can't compete with your irrational invective and sheer hostility. Perhaps you should get help with that?...

I won't close this Issue until it has been resolved to my satisfaction.

[neyoneit]

"Get help with that." Sure, coming from the guy who called me a cheapskate for filing a bug report.
"I won't close this until it's resolved to my satisfaction." Interesting, because "two minutes" ago you said you don't see much point in continuing. Make up your mind. 😂

[rlktradewright]

Wow, you really do have a problem!

I meant that there's not much point in continuing arguing with you. We are not going to see eye-to-eye..

[neyoneit]

Back to "no point in continuing" now that you've run out of arguments? Funny how that works. I imagine you're just as pleasant to deal with in real life.
Again, feel free to close this issue, but you might end up unsatisfied. 😂

[perriea]

Hello,

In case you are not aware, some of us who follow this project closely (as it is essential to us) are notified of all discussions on it.

@neyoneit stop. He owes you nothing. This is his project, he is almost single-handedly maintaining it, and he does so for free. The level of hostility in this thread is unacceptable regardless of what was said first.

@rlktradewright consider adding a Q&A section to the wiki (covering common cases like unfunded paper accounts, autorestart limitations, etc.) and linking to it in the issue template. You could also configure the template to explicitly ask reporters to confirm they have read it first and feel free to close issues that clearly haven't. I am happy to open a PR for both the template and the wiki page if you would like.

[neyoneit]

@perriea Why? For the record, I was perfectly civil (check all my posts, even in the previous issue I thanked him many times) until words like "cheapskate" started flying and bug reports started getting dismissed based on account funding status. I simply returned the energy he set. He dug that hole himself and has been shoveling ever since.