Stake in the ground, as the sister post said, non-corroding, in the simplest case.
But there is more to it: The stake has to have permanent contact to some electrically conductive layer in the ground, so you need to take geology and local climate into account. In central europe, with generally wet climate, you just need to reach the year-long stable, frost-free, local water table at a depth of (usually) between 1m and 10m. If you cannot reach sufficient depth, don't know the required depth, a simple stake isn't going to cut it. Because in case of an electrical fault, the grounding has to withstand and dissipate in the order of a few hundred Ampere. To achieve that you then shallowly bury lines of non-corroding material in a grid, or bury a grounding net something like 1 to 2m deep over an area of 100m^2 to 10000m^2.
If you are on sandy or rocky ground, permafrost, arid climate and no handy body of water is nearby for grounding, you need to have a far larger grounding net or use conductivity-enhancing methods like permanent watering, adding salts or carbon to the soil or replacing it outright with something more conductive. In all, very expensive.
And as for large installations, you just measure the soil conductivity, calculate the necessary grounding current and scale up the aforementioned methods.
This must be a European design. In the USA, governed by NFPA 70 - The National Electric Code, the ground rod's purpose is to establish the ground voltage reference to reference the electrical systems of a building to it and to dissipate any charge buildup on the circuits.
The ground rod is not there to carry current to interrupt a fault to "protective earth" - the green or green/yellow. The circuit breaker interrupts a short as you bring the protective earth wire back to the main disconnect of a building where it bonds to the neutral wire.
If a fault occurs, an unregulated amount of current flows on the protective earth wire to the breaker panel and a circuit breaker interrupts the circuit.
I'm no expert, but here in Norway we predominantly have IT systems[1], though new installations are mainly TN.
In the IT systems, the protective earth is not bonded to the neutral. Thus in case of a fault, the protective earth should be low impedance to ground so that the circuit breakers trip. At least that's my understanding.
Technical nitpick, if it detects earth leakage and trips based on that, it's an RCD, not just a circuit breaker. Otherwise yeah, there's different approaches to earthing - in Australia for domestic stuff we have mandatory RCDs which work as you describe above, but then also in industrial/mining settings we have the big green/yellow cables which will directly sink current (potentially hundreds of amps) to ground, hopefully stopping you from getting bitten.
A thermal breaker won't trip on leakage current or arc faults either. I was trying to describe the classic "dead short" that will trip a thermal circuit breaker or fuse without the circuitry for arc or leakage detection.
Ground water isn't really receding, it's just pumped to a low level to benefit agriculture at early spring, which then fucks everyone up if the spring and/or summer is dry.
Usually you can just tie it to an outside copper plumbing pipe, it's metal and makes good contact with your local ground.
Disclaimer: Not intended in any way as professional electrical advice yada yada. Just what I've read.
(Also all sorts of weirdness can take place around grid earth vs. local earth, eg. during thunderstorms. Earthing is its own entire engineering discipline. :S )
You need to put as much as needed to achieve 4 Ohm to ground or less. This can be up to ~ 10 stakes at a couple of meter interval, it depends a lot on the soil type.
What does this mean? My naive reading would presume a stake in the ground?