Kerbal Space Program

In the example I gave above there is an engine both in the center stack and on the radially-mounted (outside) stack. If there is no engine in the center, however, the bug isn't present.

The fuel line is pointing inwards. The fuel flows from the top outside tank to the center tank, keeping it full until the top outside tank is empty.

Some additional observations:¶

I've reproduced this bug in both 0.19.1 on Mac OS X and 0.18.2 on Windows. I forgot that I hadn't yet updated the game on my Windows machine until I saw that you had marked this bug as Windows-only. Additionally, I can rule out add-ons as the Windows machine is 100% stock.

The bug will present itself if you place an additional fuel line from the center tank to the upper tank (creating a loop between the two) even when there's no engine below the center tank. So essentially, fuel only has to be flowing through the inward fuel line for the bug to be triggered.

The outside lower tank seems to be draining slower than I would expect. I'm suspect that the outside engine is actually drawing from both of the tanks above it.

I don't think it's possible to make that determination without running this setup in a debugging environment.

That said, is that even possible? I don't know of any situation where fuel will flow upward.

Is this even a bug ? From the central engine's point of view, is the lower radial tank the farthest upstream tank and is thus drained first.

That could certainly be the case, though I'd still classify it as a bug. Fuel shouldn't ever flow upward unless it's via a fuel line.

I don't see this as a bug too, because the fuellines define the last tank that will be drained. So if you don't want the upper tanks to be drained last don't attach the fuellines there.

Also if the fuellines don't drain their attached tank last what do you expect them to do instead?
Be attached to the upper, draining it like normal first, then when the upper is empty ignore the lower because their attached tank is then empty.

In general I would expect one of two behaviors, and only one in the the current design of KSP:

1. Since tanks drain downward, and are physically separate, fuel shouldn't flow upward unless there is a overt reason. A fuel line implies that there is some pumping mechanism.
2. Stacks of tanks should combine into one logical tank (which would then drain all as one tank). I vaguely recall a mod that accomplished this.

Considering the second expected behavior, perhaps this would be better as a feature request.

The reason this is an issue is because I was trying to build a spacecraft that would partially drain its central core before the outer stages had completely drained. Specifically, I wanted to mimic the behavior of Falcon Heavy, which will (eventually, once cross-feed is implemented) power three of the main core's 9 engines using that core's own propellant. To mimic this behavior I attached a fuel line to one of the higher tanks on the outer rocket cores, but the result was that the outer core engines were starved of fuel (due to the bottom tank draining into the central core) long before their tanks were empty.

This is a well understood and well documented (here: http://forum.kerbalspaceprogram.com/threads/64362-Fuel-Flow-Rules-%280-24-2%29) consequence of the "fuel flow" mechanism.

Eh, at this point I'd rather just leave it as-is. The behavior is certainly unexpected, but I'd rather make a feature request to have stacked tanks merge together.

Testing the original craft in 1.0.5.1028 the following observations are made.

1. Fuel is drawn from the topmost tank first in the case of the none fuel line attached stack.
2. Fuel is seen to be drawn from the bottommost tanks of the fuel line attached stacks as well as the topmost. Middle remaining fullest.
3. Stacks with fuel lines are depleted first.
4. Detaching the fuel lines leaves all stacks being depleted top down.
5. Enabling decoupler crossfeed without fuel lines, sees the centre stack use its fuel before the outer stacks.

The behaviour of the fuel flow within a stack was changed in 1.0.5 to help with keeping the CoM low down on the craft by drawing from the furthest tank from the demand (Usually top down), as well as the addition of the ability to control crossfeed on decouplers.

With this information, I draw the following conclusions.

1. The topmost down for none fuel line attached stacks is correct.
2. Fuel lines supplying the centre stack are the new demand for the outer tanks they are attached to. The furthest tank is now the bottom of the stack. Fuel is also being used by the engine from the top.
3. The fuel line attached stacks are supplying both engines and the centre stack, so therefore empty first.
4. Detaching fuel lines appear to make the stacks behave as expected. Topmost down.
5. Enabling crossfeed shows that the centre stack is depleted first. It has two engines, so therefore higher demand.

According to the new rules, and my observations, the behaviour is now in keeping with expectations. With regard to the middle tank on fuel line attached stacks, fuel is being demanded from both ends of the stack due to the fuel line becoming a demand from the top, and the engine demanding from the bottom with both emptying the furthest tank in turn, therefore it remains fullest. This can be confirmed by removing the engines from the fuel line attached stacks and observing the flow from bottom to top on those stacks.

The only anomaly that I see is a new behaviour where the crossfed centre stack is flaming out before the outer stacks. This could be due to the double engine demand or the outer tanks taking precedence on any remaining fuel in the centre single tank. A freshly built craft in 1.0.5.1028 exhibits the same behaviour.

To satisfy the expectations of the original report. Attach the fuel line from the bottom of the outer stack to the centre stack. Fuel will then be drawn from the furtherst, or top tank first.

Confirmation of these observations, and satisfied expectations are welcome.

This was resolved (and then some) with the addition of the tank prioritization feature in 1.2.