That makes sense, rfedoruk. Even if the workflow is finished, there is a legitimate business need to keep the workflow around until at least after the task is closed out. Here's a customization you can do to only clean wf_context records whose target record has been closed out. It works for tasks and task_sla records. It will only delete workflow contexts whose related task has been closed or whose related task_sla record has ended. I think that covers all use cases for workflow contexts. I have tested it out on my demo instance pretty thoroughly but please do your own testing to be on the safe side. It should be fairly performant, provided you have a database index on task.closed_at and task_sla.end_time. If you do not have these indexes I recommend adding them.
Robert, if you have not already, can you please open an incident with the HI team and reference this community thread and your other community post? We should open a Problem record in HI to track the issue. Given that the sys_auto_flush record comes back in Jakarta, I suspect that new customers will run into the issue that Robert described. I cannot be certain that just deleting the sys_auto_flush record is a safe option. However, I did not find any incidents in HI where this was found to cause performance impact and none of my colleagues were aware of this being a problem.
Alter (or create) the wf_context record as in the below screenshot:
Create a global Script Include named "getWFContextID" with this code:
var getWFContextID = function(seconds) {
//get all sys_id's of tasks that closed in the last 2 hours (i.e. since the last time the table cleaner ran)
try {
seconds = seconds || 15552000;
var secondsAgo = new GlideDateTime();
secondsAgo.addSeconds(seconds * -1);
var result = "";
var relativeTime = "RELATIVEGT@hour@ago@2";
//grab all task records closed in last 2 hours
var taskClosed = new GlideRecord("task");
taskClosed.addEncodedQuery("closed_at"+relativeTime);
taskClosed.query();
while(taskClosed._next()) {
result += "," + taskClosed.sys_id;
}
//now grab task_sla records
var slaClosed = new GlideRecord("task_sla");
slaClosed.addEncodedQuery("end_time"+relativeTime);
slaClosed.query();
while(slaClosed._next()) {
result += "," + slaClosed.sys_id;
}
var wfIds = "";
var workflowContext = new GlideRecord("wf_context");
workflowContext.addQuery("id", "IN", result.substring(1));
workflowContext.addQuery("ended", "<", secondsAgo);
workflowContext.query();
while(workflowContext._next()) {
wfIds += "," + workflowContext.sys_id;
}
return wfIds.substring(1);
} catch(e) {
gs.log("getWFContextID failed: " + e);
return null;
}
};
I suppose this code might be better re-written to query all wf_context records over X age and then query their related task/task_sla records to see if they are still open but I would worry that that result set might start to get pretty large. The way I wrote it we only grab task/task_sla records that closed in the last 2 hours - should remain a fairly small, constant result set size.
