I've been experiencing strange request spikes from a few referenced, branch versioned feature services lately and I'm trying to figure out if it's normal behavior or if something is wrong.
Here's an example: Earlier today one of my colleagues tried making an edit using one of the services but the changes didn't take due to an underlying DBMS error (per Server logs). Around this time that same service jumped from a few hundred requests to almost 12,500. Is that normal request rate activity for a failed edit? I've also seen this kind of behavior from another, similarly configured service, but often at the beginning and end of the day--I assume this correlates with staff opening up edit spaces at the start of work and then pushing edits as they wrap things up, respectively. For that service I'll often see request rates spike to around 20,000.
During these times the services, running on dedicated pools since they're branch versioned, will also tend to max out their number of allotted instances, one of which is set to 13 to match the number of processors on the application server +1. These services are all password protected and I seriously doubt that many human users are actively connected.
Support hasn't been able to find a smoking gun as of yet, and I can't find much in the way of community discussion when it comes to request rate spikes. Anyone have thoughts?
