In the comments you say
power rails are +-12V...
when I blowed on the IC, frequency dropped and when I left it running
for 10 minutes, turned it off and then on again after 10 seconds, it
was showing the same frequency but when left for some time and turrned
on again, it showed frequency around 1K3 Hz.
This is a sure sign that self-heating is causing at least some of the drift. The LM347 has a typical quiescent current drain of 8 mA. 24 V x 8 mA = 192 mW. Junction-to-ambient thermal resistance is ~43 °C/W for the DIP package and ~75 °C/W for the SOIC version, so there will be significant temperature rise inside the chip. This can cause various parameters of the IC such as input offset voltage, gain, phase shift and output saturation voltage to drift. Heat may also flow into other components in the circuit and make them drift too.
An oscillator that doesn't have level stabilization will increases its output level until the waveform becomes compressed. In this case the first op amp's output will hit the supply rails (less saturation voltage of 1.5 V per rail), producing a square wave that is low pass filtered by the subsequent stages. The last stage's output is close to a sine wave, but the amplitude is too low. You could lower the gain until the first stage only barely saturated, but then the level would become very sensitive to changes in temperature and supply voltage.
For your application it might be better to produce a square wave with a separate low drift oscillator (eg. LMC555 timer or CD4060 oscillator/divider), and use the LF347 as a 4 stage low pass filter to suppress the harmonics.
