On Wed, Apr 17, 2019 at 12:32:16PM +0000, Suresh Thelkar wrote:
> Summary: It looks like only one index got corrupted and reindex on the
> particular index indeed solved our issue. We tried to reproduce this issue 4-5
> times and all the time only one index i.e. pbs.job_attr_idx: got corrupted. So
> given this circumstance can we rely on reindex of corrupted index or at whole
> database level as a SOLUTION to the pg_upgrade that we are seeing.
>
>
>
> It appears that pg_upgrade corrupted an index of an otherwise healthy source
> database during the upgrade process. Seems like pg_upgrade bug to me and could
> we know if it corrupts anything else ? or in other words what should be fixed
> in pg_upgrade.
My wild guess is that the index is corrupted on 9.3.12 but only detected
as corrupted on 9.6.12. Can you try running reindex before doing the
upgrade to see if it works after the upgrade? If that is wrong, it must
be something else since pg_upgrade never modifies files when
copying/linking them between versions.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ As you are, so once was I. As I am, so you will be. +
+ Ancient Roman grave inscription +
