How do I create a passive health monitor?

Just curious what you mean by "passive" health monitor...

Version 9.4 is supposed to have passive monitoring features, not sure how that ties into iRules though

LocalDirector had a nice passive monitoring process whereby it took a server out of service when X consecutive syn packets went unanswered. I wish the LTM supported this on the monitoring side. You can affect an LB_FAILED event and reselect based on your tcp profile, but this tuning should be available as a simple monitoring scheme.

I found this from the LTM 9.4 release notes... I can't find any other documentation. F5 consultants were onsite and mentioned passive monitoring as a possible solution to my problem.

 

 

 

"The BIG-IP system now includes a feature known as passive monitoring. With passive monitoring, a pool member can be marked down sooner than the customary three successive bigd health check failures. Implementation of this feature requires the use of the iRulesTM feature."

 

I just opened a case with F5 support on this same issue. Once they point me to the documentation for this feature I will post the info here.

Basically what this involves is looking at the HTTP status codes that come back in the HTTP_RESPONSE event and invoking LB::reselect if it's a 404 or 500 or something like that. I believe LB::reselect is new in 9.4.

 

 

In the http profiles you will also notice a new field in 9.4 that does a fallback on whatever response codes you define (it's essentially a canned iRule along the lines of what I described above).

 

 

Denny

EDIT: sorry duplicate post (stupid slow EVDO connection..)

Hi Dave -

 

 

LB::reselect was introduced in one of the 9.2 releases. I know it is available in 9.2.3 or better. LB::reselect can be used in LB_FAILED, which is only triggered if the selected pool member doesn't respond. It cannot be used in *_RESPONSE events if the server does respond.

 

 

If the server does respond and the response indicates a new node should be selected, for HTTP traffic I believe you could use LB::detach and HTTP::retry in HTTP_RESPONSE event to send the request to a different pool member. I don't think there is an equivalent command for replaying a TCP request.

 

 

Passive monitoring in 9.4 is specific to HTTP response codes.

 

 

HTH

 

/deb

Passive monitoring in 9.4 is specific to HTTP response codes.

 

 

 

...Bummer. I was hoping the passive monitoring would be more feature rich than this.

Where is passive monitoring enabled/invoked? I need to mark down nodes as soon as a server error occurs. Waiting for a regular monitor takes too long.