How to detect replication doesnt work and get noticed?

I want to get notice by netsend, email or any other methods
when replication doesn't work or malfunction.
Does anyone know how to do this or any other solution you use for this kind of issue?
Thank you..Every agent that runs your replication (snapshot, logreader, distributot, and queuereader) have associated SQLAgent jobs. All you have to do is modify those jobs by going to Notification tab of the job properties and selecting notification method you'd like to get. You should have an appropriate operator set up before, or you can do it right there by selecting New Operator.|||Thank you for reply..

The main problem I have is how to find out if the replication doesn't work..
I can notice with agent history or icon.

Sometimes the icon become X marked, some times agent keep trying to connect and generate error with forever loop, sometimes gives I/O error while working..

How to find out these and what's best way to find out if replication doesn't work properly?|||Have you looked at what I posted about?|||Yes I did..
I knew that part..
Could you give me more detail with example with specific error cases?
Then, I understand more clearly.|||see attachment|||Thank you for the image..
I think I got it at this level..

I just wonder is there a way to run a custom application when job fails rather than sending Net-send..

Since I have to check hundreds of systems replicated, if I use net-send I will get too much of it and hard to maintain.
If I can run my custome application reporting problem organized, it will be easy to maintain.|||In that case you should create a batch file and call it from task window .

In SQL Scheduler when you click on the task and click advanced it says

" On failure action: and then there is a drop down . All you have to do is add another task to the existing job and call it here

It will say

"On failure Action : Goto Step #2" In that step #2 you can all the windows batch file .. (which can contain anything like ISQL, another program ...)|||Thank you!
I got it!|||Another way of doing it would be to create a custom alert and a scheduled task. When conditions for the alert are met you can invoke that scheduled task that may contain ... see previous post.