Forum Discussion

lipos_54863's avatar
lipos_54863
Icon for Nimbostratus rankNimbostratus
Aug 28, 2009

monitors behavior

Hi,

 

I got a strange monitor problem.

 

One HTTP VS has one Pool assigned to it with 4 Node. Configuration as simple as it gets.

 

At first I assign default http monitor with send string of GET /

 

Couple weeks after I created new monitor that was doing GET / HTTP/1.1

 

Three day ago I created another one doing GET /lb/lb.php HTTP/1.1\r\nHost: xxx.com\r\nConnection: close with receiving sting of UP-AND-RUNNING"

 

Everything is working well and only the last monitor is assigned to the Pool.

 

Hoverer, after running: tcpdump -ni VLAN01 -X dst host IP

 

I see:

 

 

HTTPGET /

 

HTTPGET / HTTP/1.0

 

HTTPGET /lb/lb.php HTTP/1.1

 

 

It's like three monitor are still running, but only one is assigned to the Pool.

 

I don't see this kind of behaviour in other pools.

 

Any idea why?

 

 

I'm on LTM 9.6.4
  • After checking the instances of http monitor and upgraded http monitor they are still there.

     

    Is that a bug?

     

    I'm thinking about manually deleting unneeded instances is that safe in this scenario?
  • hoolio's avatar
    hoolio
    Icon for Cirrostratus rankCirrostratus
    Hi Lipos,

     

     

    Do you have any node monitors applied? You might try running qkview to get a snapshot of the current config before testing anything more. If you list the config running in memory using 'b list' do you see the monitor name for the 'GET / HTTP/1.0' monitor associated with the pool or individual nodes? If you reload the config using 'b load' does the issue continue?

     

     

    Aaron
  • I confirm that only on monitor is assigned to the pool: monitor all min 1 of http_check_BC16

     

    I don't have any monitors applied to a nodes - I'm only assigning monitors to the pools.

     

     

    I'm updating active lb and checking passive every time - 'b load' is done every time I'm syncing my config.

     

     

    As mentioned before, I still able to see the instances of the server assigned to monitor1 and monitor2 even though monitor3 is used only.

     

  • hoolio's avatar
    hoolio
    Icon for Cirrostratus rankCirrostratus
    That's odd... if you've made the change to the pool's monitor definition, sync'd the config from the active unit to the peer and verified no other monitors are added to the pool, you obviously shouldn't be seeing the old monitor requests being sent.

     

     

    You could enable debug on the monitoring daemon, bigd, and check the logs for more info. You can enable bigd debug (b db bigd.debug enable / disable) and check the log file (/var/log/bigdlog). Make sure to disable the logging once you're done testing as it's quite verbose and will eat up a lot of disk space.

     

     

    Regardless, it would be good to open a case with F5 Support to get help determining if this is a config issue or a bug.

     

     

    Aaron