It has been (unforgivably) long since my last blog post so let me try and rehabilitate myself by posting a couple of recent interesting investigations. I’ll start with one that combines two of my favourites: stack profiling and “twittershooting” (collective troubleshooting using Twitter as a collaboration tool). The database in question is a 3-node RAC cluster, Oracle version 18.11 being upgraded to 19.10. The problem was detected during the pre-upgrade testing with “compatible” parameter set to 18 (that part was essential for the backout plan in case of major post-upgrade complications). I mention this because for whatever reason, the problem only showed up when the parameter was set.