I told you twice

8 July
I would bet 50c that your logic on driving the MOC3021 is inverted and you are turning it off when you think you are turning it on.

9 July
I repeat: I'll bet you 50c that P1.3 is driving the cathode of your MOC3021, (provided you are indeed driving the LED via a resistor and connected to your Vdd).

Yet you don't want to listen. Now we have your schematic and you owe me 50c.

When you execute SETB RELAY, you think you are turning the triac on. You are not. You are turning it OFF. The triac only turns off at a zero crossing point. When the cycle starts the triac is on (P1.3 is low). Somewhere in the cycle you turn it it off (P.13 goes high) and then a millisecond later you turn it on (P1.3 goes low). Since this off/on transistion does not happen at a zero crossing point the triac does not turn off ever.

If you had tried my 3 step suggestion ( in my 9 July message http://www.8052.com/forumchat/read/177201 ) you would have discovered that.
The simple solution is to change every SETB RELAY to a CLR RELAY and vice versa.

This may not be your only problem. Go through everything that has been written in this thread, perhaps your design will become worthy.

You need to see whether the ~3.8mA that you pass through the MOC3021 is sufficient to guarantee to turn it on.
Also your zero crossing point is not at zero, but delayed by a fixed amount. The input voltage must increase to at least 1.8V (3 diode drops) before the first transistor turns on.

I don't know why I bother to even tell you this since you have not paid any attention to anything I or anyone else has said either. I'm done with trying to help.