RE: [SA-list] Suggestion: recheck dependent tests on second-knock

[Robert Haerkens / Native networks]

Title: Bericht

Okay, but I don't agree with that 100%. If SA gives just a view of the situation at the time of the check you won't have to do a second knock at all. If the entry appears to be down, that's the way it is at that time. Maybe, the test-packets are lost so you want to do a second knock. Then I say: maybe a depending entry went down, give them also a second knock.....

 

I understand that it is the nature of SA to test one entry after the other, determine the state, trigger an alert and go to the next entry. So it is difficult to go back. But maybe the possibility to group tests could give a solution to this. You could test the whole group (en second knock the whole group) and trigger alarms after the complete group is tested or re-tested. Or, why not first do all tests and then at the end of the cycle to all alarms, website updates, statistical logging, etc.

 

This is the last thing I will write about it. I really like SA and think you do a great job Dirk. I just wait and see it evolving...

 

Robert

 

 -----Oorspronkelijk bericht-----
Van: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Namens Dirk Bulinckx
Verzonden: dinsdag 19 augustus 2003 21:26
Aan: [EMAIL PROTECTED]
Onderwerp: RE: [SA-list] Suggestion: recheck dependent tests on second-knock

However you change it you will always have a possibility were you "just missed" the down of the entry above the entry....that's just the way it is.  The checks are always a view on the situation at the time of the check, not just before or just after it. 

 

dirk.

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Robert Haerkens / Native networks
Sent: Tue Aug 19 8:38 PM
To: [EMAIL PROTECTED]
Subject: RE: [SA-list] Suggestion: recheck dependent tests on second-knock

Dirk,

 

The way SK is implemented now works fine when servers went down before the test-cycle starts. It is very likely that a sub-test fails because the test where it depends on went suddenly down. So, you will get a lot better diagnostic when SA does one step back in case of a "possible down". It can discover if the problem is larger.

 

With SA it's almost possible to identify the problem exactly when you do the right tests and make them in the correct order depending on each-other. To test a mailserver I would first telnet it's Internet router, then try to connect to the SMTP port and then try to use SMPT2POP3. If one of these tests fails it would be very usefull if SA could do a step back in dependencies to check out the exact problem.

 

Thanks for your time,

 

Robert

 

 

-----Oorspronkelijk bericht-----
Van: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Namens Dirk Bulinckx
Verzonden: dinsdag 19 augustus 2003 18:58
Aan: [EMAIL PROTECTED]
Onderwerp: RE: [SA-list] Suggestion: recheck dependent tests on second-knock

Servers Alive will recheck those that depend on the SK item, that's the right way to do this, not the other way around.

 

dirk.

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Robert Haerkens / Native networks
Sent: Tue Aug 19 5:43 PM
To: [EMAIL PROTECTED]
Subject: [SA-list] Suggestion: recheck dependent tests on second-knock

I have a suggestion but I don't think it is very easy:

 

Would it be possible to change the behaviour of the "second knock" function so it "re-tests" not only the failed test but also the test where the failed test depends on?

 

We use SA to test over the Internet. Before we do tests we first check if our Internet connection is functioning (if that is not working, you dont have to do other tests since all other tests depend on our Internet connection. This works perfect in theory but in practise when our connections drops (since I use SA I know it drops far more often then the ISP tells me), it always drops in the middle of a test an not exactly before we do our "self-test".

 

Maybe the easiest way is to change the Second Knock to retest the goup in the second knock in stead of the failed test?

 

Robert