On Wed, Mar 21, 2018 at 4:29 PM, Peter Geoghegan <pg@bowt.ie> wrote:
On Wed, Mar 21, 2018 at 1:38 PM, Jeremy Finzel <finzelj@gmail.com> wrote: > A server restart and upgrade to 9.5.12 (at the same time), as expected, made > the issue go away. Still doesn't give us any answers as to what happened or > if it would happen again! Thanks for the feeback.
You may still want to use amcheck to look for problems. The version on Github works with 9.5, and there are Redhat and Debian pgdg packages. See:
The "heapallindexed" option will be of particular interest to you - that option verifies that the table has matching rows for a target index (in addition to testing the structure of a target B-Tree index itself). This is probably the best general test for corruption that is available. There is a fair chance that this will reveal new information. -- Peter Geoghegan
Thank you for the recommendation. I ran both amcheck functions on all 4 indexes of those 2 tables with heapallindexed = true, but no issues were found.
