How to validate my glm model probably the title is not very clear but here goes :
a built a gml model on my train set with 
model=glm(y~x1+x2)

and now i'm predicting the output on the test set with      
predict (model, newdata)

how can i know if my predictions values are "acceptable" or not comparing to the real ones ? nomaly there is an error prediction ? 
in anothers words, how to know if my model is "good "and that it predict very well with a certain error ?
Some data : 
x1="begin", x2="time", y="final"
29180          523      29757
29289          439      29739
29413          339      29811 

 A: You can use mean-squared error in this regard. It helps to know  how much erroneous your model. 
Here is R code- 
pr.lm <- predict (model, test) # just assignment
# test set is to be predicted, test$y : the dependent variable of model
MSE.lm <- sum((pr.lm - *test$y*)^2)/nrow(test)

Hope it will help.
A: When you mentioned test set, I guess this is the data without actual result.
But to validate the model you've built, you need a hold-out sample which has actual result on it as well.
For the validation purpose, it would be ideal to set a hold-out sample from your training dataset as validation datset. Normally, I would choose 70% of the training dataset for modelling process, and rest of 30% of the training dataset is for validation. Of cause, you need to balance this proportion in case there is no enough data in your validation set.
Then, the validation process is conducted on this hold-out sample that compares model prediction with the actual result.  An example is like: https://stackoverflow.com/questions/21380236/cross-validation-for-glm-models-in-r
