A few things to watch out for in the 749x series ripple counters:
(1) the counter is split into a ÷2 and a ÷8 (7493), ÷2 and a ÷5 (7490) or ÷2 and a ÷6 (7492) counter and you must take the output of the first ÷2 Qa and feed it to the CLK input of the second section and,
(2) the resets on these counters are active HIGH rather than active LOW. If you let them float, the counter will reset; in addition, the 7490 has reset-to-9 inputs which must also be held LOW to allow the counter to count.
(3) the power supply pins of these particular counters run counter to the "normal" pin locations on TTL chips. The 749x series were some of the first TTL chips designed and as such, did not fall into a standardization of supply pinouts, proably because at the time, because of where the supply points were on the die, it was the only way to get power in without crossing bonding wires or totally redesigning the chip.
You may notice that all TTL ripple counters all count on the negative-going edge while all TTL synchronous counters (and almost-synchronous counters such as the 7419x series) count on the positive-going edge. That's why it's dicey trying to cascade TTL ripple and synchronous counters.
Now, if you've taken care of the above two quirks and your counter is still counting to "2" and locking up (or repeating the count), you may have power supply glitches getting in. Be sure to bypass each counter AT THE POWER SUPPLY PINS with a 0.1µF disc ceramic capacitor, one for each counter. It's amazing how a lack of supply bypassing can change how clocked logic operates!
Dean