Forum Discussion
Multiple Monitor Timers
- Jul 16, 2015
Strictly speaking, the external monitor could change the timeout value on itself by executing
tmsh-- or an iCall periodic script could run the monitor page check and modify the timeout value -- but I strongly recommend against doing that, or ever having an external monitor change configuration, for that matter. The reason is, if the BIG-IP is part of an HA cluster, it will make the member out-of-sync (and if autosync is turned on, it may push the config when somebody is making changes on another unit, wiping out whatever changes were made).An external monitor could solve this in a number of ways, but here is logic for one way to do it:
- Run the test that was being performed by the monitor (e.g., if the monitor if http, use curl to retrieve the page and parse the result);
- If the test in 1 succeeds, return an "up" value (that is, print something to stdout), remove the file cited in 5 (if it exists), then exit;
- If the test in 1 fails, run the test against the monitoring page;
- If the test in 3 fails or if the results do not indicate that maintenance is happening, remove the file cited in 5 (if it exists), then exit without writing anything to stdout (this will trigger a "down" status);
- If the test in 3 succeeds and indicates that maintenance is happening, check to see if there is file named /var/tmp/mymonitor.$ip:$port.tmp ($ip and $port are always the first two command-line argument provided to the monitor);
- If the file does not exist, grab the current unix epoch time in seconds and write it to the file, return an "up" value, then exit;
- If the file does exist, read the value, grab the current unix epoch time in seconds and compare them;
- If the delta in 5 is less than your second threshold (30 seconds), return an "up" value and exit;
- If the delta in 5 is greater than or equal to your second threshold, exit without writing anything to stdout (which triggers a "down" status).
Here it is in psuedo-code(ish) ($1 and $2 are the command-line arguments supplied to the script):
all external monitors are automatically passed the target IP and port as the first two positional parameters. If the IP is IPv4, it will start with ::ffff:, followed by the IPv4 address as dotted-quad (e.g., ::ffff:10.10.10.1). ip=`echo $1 | sed 's/::ffff://'` port=$2 file="/var/tmp/$0.$ip:$port.tmp" monitor_page="https://my.monitor.example.com/status" first_test=`curl -Lis http://$ip/ | grep -E '^HTTP/1.1 200' if [ "$first_test" != "" ]; then rm -f "$file" echo "UP" exit 0 fi assuming that I'm looking for the string "In Maintenance Mode" second_test=`curl -Ls $monitor_page | grep 'In Maintenance Mode'` if [ "$second_test" == "" ]; then rm -f "$file" exit 1 fi if [ -f "$file" ]; then last_time=`cat "$file"` curr_time=`date +%s` else date +%s > "$file" echo "UP" exit 0 fi if [ $(( $curr_time - $last_time )) -lt 30 ]; then echo "UP" exit 0 else exit 1 fi
Firstly, why would you not want the pool members to be marked as "down" after the restart? Since the targets are not accepting connections, presumably you don't want to forward to those pools. If the BIG-IP receives a RST from the targets, it will essentially proxy that RST to the client. I assume you have a good reason for this choice, I'm just trying to better understand the situation, and this part is a bit unclear to me.
You distinguish between two different responses ("X" and "Y") based on the conditions for the pool members. How do those responses differ?
I'm guessing that your only recourse here is to use an external monitor, which can really do just about anything you want. You would, however, need to write it yourself (or engage F5 Professional Services to assist you in this endeavor). The following links provide additional information about external monitors:
- Ryan77777Jul 16, 2015AltocumulusI don't want them marked as down because they're not "down". They're just really slow because the config guys made an app change that recycled all of the members at once (of which I/they have no control over). However I do have control over the F5... so if I could switch timers depending on what the F5 receives back from the monitoring page (text on the page, whatever), then I could say "hey, they're making a config change right now according to the page I'm monitoring, wait for 30 seconds to hear a response back instead of 10" or the like. How would an external monitor work? The timeout value isn't controlled there as far as I can tell is it?
Recent Discussions
Related Content
* Getting Started on DevCentral
* Community Guidelines
* Community Terms of Use / EULA
* Community Ranking Explained
* Community Resources
* Contact the DevCentral Team
* Update MFA on account.f5.com