> Does that error persist if you delete all files named pg_internal.init
> in the data dir? (those files are a cache of the system tables and are
> regenerated on the next new connection if they are missing)
There were a few of those files, so I removed them:
$ find data -name pg_internal.init
data/base/1/pg_internal.init
data/base/16401/pg_internal.init
data/base/12292/pg_internal.init
data/base/12291/pg_internal.init
data/base/16402/pg_internal.init
data/global/pg_internal.init
$ find data -name pg_internal.init -exec rm {} \;
$ find data -name pg_internal.init
$
Starting up postgres again and querying pg_database still gives the
same cache lookup error.
Another person noticed that the missing directory
(pg_tblspc/1932420460/PG_10_201707211/16400) corresponded to an
existing directory under data/base/16400, so he copied the contents
from data/base/16400 into pg_tblspc/1932420460/PG_10_201707211/16400
(making the previously non-existent parent path
"pg_tblspc/1932420460/PG_10_201707211/" along the way). Doing that
allowed the main database to once again be queried, and all the data
appears to be present. That system still can't query its pg_databases
table, so I assume it's still pretty badly broken, but at least we can
try a pg_dump and recovery onto a new, clean, system. I still have my
copy of the broken database though, and I'd love to know if there's a
proper fix for it, so I'll keep this discussion going as long as
anybody has ideas :)
