The shortest time between executions of a scheduled agent is 5 minutes. But very often you want to run some important function with only 1 minute interval. Your function takes maybe only 2 seconds to run, but you still are required to wait 5 minutes before it can be started again at the next agent invocation.
The agent I created makes it possible to simulate running agent every 60 seconds, or even every 2 seconds if you wish so. Some restrictions apply :)
It uses the fact that Notes counts 5 minutes from the START of the previous instance, not from the END. So if your agent takes 4 minutes 50 seconds to run, the next execution of the agent (assuming perfect conditions) will be just 10 seconds after the previous execution is finished. Delays caused by server load and other unforeseen delays can cause the agent to wait from 30 seconds to 2 minutes until the next execution. During many tests, I havent's observed execution delay longer than 2 minutes.
Even with the worst case 2 minutes delay, it's much better than standard 5 minutes delay. Note that this delay is only for the time between agent executions, within the started agent you can call the function every 2 seconds if you wish.
What if the function takes more than 5 minutes to run? Well, the next instance of the agent will wait until the current instance is finished and then start after a certain delay. In my tests the delay was almost always 2 minutes. The delay if the agent ends 15 seconds BEFORE the 5-minutes period (4m 45s) was according to my tests
always 1 minute 5 seconds. These numbers will most certainly be somewhat different on your server.
The main disadvantage of this method is that one of the Agent Manager's threads will be constantly busy, so you would need to increase the number of agent threads to make it possible for other scheduled agents to run too.
To enable your agent to run every 60 seconds, do following:
1) In your existing agent, move the code from Initialize method to MainAction method.
2) Paste the code from the agent below into the Initialize method of your agent.
3) Change variable values as needed.
Agent is configured to execute the MainAction() function every
60 secondss (runinterval), max
100 times (runmaxtimes), during the 5 minute period (agentschedinterval).
Agent execution flow. Click to enlarge.
"Gone in 60 seconds" agent:
Sub Initialize
'Created by Andrei Kouvchinnikov, www.botstation.com
Print "******************* Agent started *******************"
Dim session As New NotesSession
Dim db As NotesDatabase
Dim expectedruntime As Integer 'time which take to run the function. required only for deciding about ending the loop.
Dim runinterval As Integer 'interval (sec) between runs
Dim runmaxtimes As Integer ' max number of times the function will run
Dim runcounter As Integer 'counter of number of times the function run
Dim runstart As Long, runend As Long, rundiff As Long
Dim notimeleft ' loop must be closed now, no time left to run another round
Dim agentstart As Long 'initial time when agent started. required only for deciding about ending the loop.
Dim agentschedule As Integer 'nr of minutes this agent is scheduled to run. required only for deciding about ending the loop.
Dim dynadjust 'instead of hardcoded runtime value, use last run period for calculation of next period
Dim uselongestperiod 'instead of the last run period, use the longest period the function took to run
expectedruntime=10 ' function is initially expected to take max 10 seconds
runinterval=60 ' function is called with interval of X seconds
runmaxtimes=100 'if function's run time variates, you can limit max number of times function runs
agentschedinterval=5 'agent is scheduled to run every X minutes. From the agent properties.
margininterval=5 'nr of seconds to add to the final round. used for situations when function's execution time is 0.
runcounter=0
dynadjust=True
uselongestperiod=True
agentstart=Timer
While runcounter<runmaxtimes And notimeleft=False
runstart=Timer
Call MainAction()
runend=Timer
If dynadjust=True Then expectedruntime=Int(runend-runstart) 'dynamically adjust expected time to actual time it take to run the function
If uselongestperiod Then
If expectedruntime<Int(runend-runstart) Then expectedruntime=Int(runend-runstart)
End If
timeleft=runinterval-expectedruntime
If Int((agentschedinterval*60)-Int(Timer-agentstart))<timeleft+(expectedruntime+Int(expectedruntime/100*30)+margininterval) Then 'assumes that function can take 30% longer time to run than the last time
notimeleft=True
Print "Exit. Function will not manage to finish one more run. Computed time: "+Cstr(Now)+" + "+Cstr((expectedruntime+Int(expectedruntime/100*30)))
End If
If timeleft>0 And notimeleft=False Then
Sleep timeleft 'finished before the expected time. sleep until next execution.
Else
Sleep Int((agentschedinterval*60)-Int(Timer-agentstart))-margininterval 'sleep X seconds-margin
End If
runcounter=runcounter+1
Wend
Print "******************* Agent finished *******************"
End Sub
Sub MainAction()
Print "Triggered "+Cstr(Now)
%REM
Here goes your code
Delete demo code below.
%END REM
Randomize
sleeprand=Int(Rnd()*10)
Sleep sleeprand 'Simulates time taken by the function's operations by sleeping X seconds
End Sub
This LotusScript was converted to HTML using the
ls2html routine,
provided by Julian Robichaux at
nsftools.com.
Output from the agent. Note the 1m 5s delay between instances.
Technorati:
Lotus Notes
