We are in the process of upgrading our ECC 6.0 from EHP5 to EHP7 support stack 11 - followed by a migration to SAP HANA.
We are running on Windows server 2012 and SQL server 2014.
We have been running a number of rehersals on our sandbox, development and test systems and they have all been running quite good - of course the prepare takes a number of days - but since it can be started in due time before the actual downtime window - then this is not a problem. We have planned for downtime in a weekend starting Friday evening and the system should be online again no later than Sunday.
We started the prepare phase Tuesday and it progressed quite good - building the shadow instance - but when we got to the ACT_UPG phase we ran into serious problems. What we observed was that in our shadow instance we could see the job RDDMASGL running - and in the process overview we could see background process running program RADMASG0_C3. This program triggered four parallel dialog sessions and all of them had the symptom that they were hanging a long time in Sequential read on table Tunknown. Looking into the details of the processes showed that the system was executing SQL statement sap_get_columns_of_index and that this took a long time for each execution. We could see that in our logfile placed on \usr\sap\sum\abap\log\SAPA742EPU that on the last upgrade test which was executed only two weeks before we started the productive migration we could see that phase 004 Object activation took around two hours - but on the production system it progressed extremely slowly so after 15 hours only the first four levels had been executed our of a total of 496 levels.
We started looking for solutions - but searching for sap_get_columns_of_index only gave one hit on Google - so not a lot of known problems on this. We then looked for SAP upgrade performance problems SQL server - and after digging around for some time we found note:
2235057 - Performance issues during SAP Upgrade with SQL Server
In the note SAP tells: Stored procedure sap_get_indexes has a known error and will cause performance problems in upgrade phase MAIN_SHDRUN / SUBMOD_PREP_DDIC_740 / JOB_RUTREMINDX with the job RUTREMINDX1. The solution is provided in SAP Note 1933739. You can replace this procedure in MSSPROCS by following the SAP KBA 2253903.
This note was not a direct match on our problem - but it pointed us in the right direction, since we now thought if there could be a problem in sap_get_indexes - then there might also be a problem in sap_get_columns_of_index. So we executed report MSSPROCS and regenerated all stored procedures in SQL server and to our happy surprise this immediately changed the speed from extremely slow to fast forward.
Hope that other people out there might benefit from knowing about this issue 