I am looking to do some custom TCP monitoring with a custom application that we need to monitor with the F5. Essentially, the application's thread handling logic will hang unless a specific set of send and receive steps are performed: 1. TCP handshake is established 2. Client sends a custom string. 3. Server sends a char in acknowledgement. 4. Client sends message header. 5. Client sends message body. 6. Server sends response header. 7. Server sends response body. 8. TCP close occurs. The custom TCP monitors I can make on the F5 don't allow me to follow this kind of send and recieve logic. Is there a way to perform these steps with some kind of custom monitor?

Here is a link to something that can be done via custom shell script. http://devcentral.f5.com/wiki/default.aspx/AdvDesignConfig/HTTPMonitor_cURL_BasicPOST.html In this example they are using cURL to send information and expect a return. You could end up writing some IF ELSE statement which would ultimately result in a status that is either UP or DOWN I hope this helps CB

An external monitor might work well. But, to make the call directly, we would need to build our own executable and place it on the F5. There are some maintanence issues with doing this, so we would rather not go that route if we can help it. Any other methods we can use here?

I think I found an excellent way to create this monitor. I found that if we call these apps by passing all of the pieces at once, the application works fine. After digging a bit, I was able to find ways to embed all of the special characters that are needed for a request to the application into the Send field in a custom TCP monitor: https://support.f5.com/kb/en-us/solutions/public/4000/100/sol4186.html However, I ran into a wall that I don't know if I can scale. The send string includes ASCII null chars. These should be able to be escaped as: \x00 However, it seems the BIG-IP web interface doesn't allow these chars. For, I got an error when trying to add the custom monitor (see attachment). Is there some other way to get these into the system?

Per SOL4186 I don't think you'll be able to use a null in the send or receive string: "Note: The BIG-IP system does not allow the ASCII character 0 (NUL)." I tried adding this in both \x00 and \x000 format through the GUI, the CLI and the bigip.conf file but all attempts resulted in the 'embedded zero' error. You could either check with F5 Support on why this restriction is in place or use an external monitor to send the null. A standard TCP monitor would be simpler and more efficient, so it might be worth checking with F5 on this. Aaron

I have contacted Tech Support. We will see what they have to say. I'll post their response here when I get it. Thanks for all your help so far.

Custom TCP Monitor | DevCentral

26 Replies

hooleylist
Cirrostratus
Sep 04, 2009
Hi again,

I am actually running into this issue myself now where I'd like to be able to use a null character in the send and/or receive string for a TCP monitor. Do you mind sharing your case number so I can pass it along to F5 Support as a reference for my case? Mine is C569837.

You might want to add the PID checking to ensure only one instance of the monitor per pool member is running at a time. It's more of a best practice than a functional requirement though.

Formulate the PID filename based on the script name, pool member IP and port PIDFILE="/var/run/`basename ${0}`.${NODE}_${PORT}.pid" If the PID file already exists, it means the last instance of this monitor is probably hung. if [ -f $PIDFILE ] then Stop the current process kill -9 `cat $PIDFILE` > /dev/null 2>&1 fi Create a new PID file echo "$$" > $PIDFILE

And then at the end of the script:

Remove the PID file rm -f $PIDFILE

Thanks,

Aaron
hooleylist
Cirrostratus
Sep 07, 2009
The bigd CR is CR41708 for allowing null characters in the send/receive strings. It's not supported now as the monitoring daemon uses nulls to terminate lines.

Aaron
Jeremy_Bridges_
Nimbostratus
Sep 08, 2009
Sorry for the delay. My case number is C565983.

Wouldn't the script you suggest incur a file I/O performance penalty? Since there is already a pretty significant performance penalty with external monitors, wouldn't I want to avoid any other potential performance problems?
hooleylist
Cirrostratus
Sep 08, 2009
The issue the PID file solves is killing a script that hangs. I think the extra overhead of the file I/O justifies the functionality. I'd guess it's not much relative to the overall resource usage of the external script and monitoring.

Aaron
Jeremy_Bridges_
Nimbostratus
Sep 09, 2009
Ok, sounds good. I will add it.

Another question for you: what is the best way to assess the performance impact of using this external monitor? Specifically, what performance stats should I look at as we monitor more pools in this way to know for sure that our BIG-IP can handle the extra overhead of external monitoring?
hooleylist
Cirrostratus
Sep 10, 2009
I'd check the top and ps output. You could try to gauge the CPU/memory usage while there is no traffic and then increase the number of pool members the monitor checks. You could then extrapolate from a few data points on what CPU/memory the external monitor uses for a varying number of pool members.

If I do anything in this area, I'll let you know. And if you try some checks, let me know how it goes.

Thanks,

Aaron

Forum Discussion

Custom TCP Monitor

26 Replies

Recent Discussions

Telemetry streaming to Elasticsearch

Error while running ansible

SSL protocol mismatch

Could not connect to SMTP host.

cat and grep command for rst messages

Related Content

How to create custom monitor

Custom HTTPS Monitor

HTTPS Custom Monitor

Custom HTTPS External Monitor

custom response page for api