We are intermittently encountering an issue with the load generated by the UWL service. The Portal is flooding the backend (ECC) with RFCs calling SAP_WAPI_GET_HEADER and SAP_WAPI_GET_OBJECTS. We go from a couple of million RFCs a day up to 15 million when these calls come in.
The impact was the very high CPU utilization on the app servers and wait time for work processes, so we reduced the logon group to only two app servers, which then breached the conversation ID limit (500) and basically brought down those two instances.
We've now opened up the logon group to the four app servers and restarted the Portal to allow balancing across them. All's now well for the moment. However, we're at a loss as to what to do next. SAP has recommended an upgrade to the service but this will take a couple of weeks of deployment and testing. We're approaching quarter end, so are in a critical period.
We've reviewed KBA 1577547 and are configured as follows:
Default Cache Validity Period, in Minutes = 5
Page refresh rate = 150
Delta Pull Channel Refresh Period (in Seconds): 60
Job in ECC (UWL_DELTA_PULL_2) = 3 mins (changed from 1 minute last week).
Any ideas?
