System 'Heartbeats'

[RSoles]

it all depends what you're trying to achieve.
I've seen systems which use ping, but that only tells you that your base OS and stack are alive.

To check full server functionality you'd need to go down the 'Phantom transaction' route, and judge the number of transactions by the amount of time you can allow to pass before you flag a dead system. 400% increase seems overkill to me.
Does the HA system allow you to the adjust the monitoring frequency?

e.g. I've seen medical systems where the integrity of other processors is checked over 10 times a second.

[SimonMac]

Just looking at a system, and the supplier puts "synthetic transactions" through to see if the other end is still there. The problem is that these transactions can't be filtered out for reporting purposes on the MI. I've trapped the 2 IP addresses and am trying to get them to filter out anything from these. So we now have a 400% increase in traffic, for no extra real user transactions.

I'd be interested if we can solve this some other technical way as well.

Whats the "other end"

If its a webserver I have just put a holding html page and pointed zabbix at it to confirm if the pages is still there.

[anonymouse]

Same here...

Just looking at a system, and the supplier puts "synthetic transactions" through to see if the other end is still there. The problem is that these transactions can't be filtered out for reporting purposes on the MI. I've trapped the 2 IP addresses and am trying to get them to filter out anything from these. So we now have a 400% increase in traffic, for no extra real user transactions.

I'd be interested if we can solve this some other technical way as well.

[original PM]

yeh we have a standard HA system spread over 2 data centres so if one dies we should know about it and fix it.

I will get more info - as I think one of the problems may be that we are only a 9-5 business and so the system seem to go into some form of rest mode when inactive.

Then when they do get a message everything needs to start up again - which can cause timeouts...

but thanks for the info most usefull

[eek]

Depends what you are using, but surely you set up 2 in parallel for high availability / failover and let their front end tell you what is happening...

I really don't see the point of trying to second guess a systems own self-healing...

[original PM]

Morning Guys....

Looking for a bit of info regarding system monitoring etc

Generally we have a few message brokers and some of the techies generally want to 'spam' the system with a message every x seconds to make sure it is still up.

There are a few reason why I do not want to this such as well spam but also unless I am using exactly the same config and routing for the 'spam' message as for the real message then it may return false results.

But generally what sort of monitoring should we have on systems to ensure items like message brokers and endpoints are up and available.

If 'spamming' is the way to go then so be it I just though there may be something more elegant.

TIA