Re: BUG #15467: The database subdirectory "pg_tblspc/1932420460/PG_10_201707211/16400" is missing.

>>>>> "tsuraan" == tsuraan  <tsuraan@gmail.com> writes:

 tsuraan> Quite a bit there, but the region around there does look weird:

 tsuraan> 00001d40  00 00 00 00 00 00 00 00  00 01 00 00 ff ff ff ff  |................|
 tsuraan> 00001d50  00 00 00 00 f4 06 00 00  01 00 00 00 6c 65 2e 73  |............le.s|
 tsuraan> 00001d60  6f 00 70 79 63 00 65 2e  00 00 00 00 00 00 00 00  |o.pyc.e.........|
 tsuraan> 00001d70  00 ff 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
 tsuraan> 00001d80  00 00 00 00 00 00 00 00  3a 00 00 ff 00 00 00 00  |........:.......|
 tsuraan> 00001d90  00 00 00 00 00 00 00 ff  00 00 00 00 00 00 00 00  |................|
 tsuraan> 00001da0  28 f8 12 00 00 00 00 00  00 00 00 00 00 00 00 00  |(...............|
 tsuraan> 00001db0  0d 00 0d 80 0a 29 20 00  00 00 00 00 03 30 00 00  |.....) ......0..|
 tsuraan> 00001dc0  74 65 6d 70 6c 61 74 65  30 00 00 00 00 00 00 00  |template0.......|
 tsuraan> 00001dd0  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|

 tsuraan> The le.s that you noticed actually looks like le.so, and then
 tsuraan> there's a pyc and an "e." after it (null-separated).

That's quite odd.

But it looks like the data here is corrupt from offset 1d5c for the next
64 bytes (+/- 4 bytes, maybe, but much more likely to be exactly 64).
That's not something I recall seeing before, but some kind of hardware
corruption issue couldn't be ruled out. I don't suppose you have data
checksums enabled?

The "cache lookup failed" error is caused by the fact that the column
_after_ dattablespace, i.e. datacl, is also corrupt (the database must
have had a non-null ACL with an additional grant to exactly one role
going by the layout of records).

 tsuraan> Also weird is that there are six different sections
 tsuraan> describing(?) template1:

That's not really weird - the pg_database rows can get updated, e.g.
from grant/revoke, and autovacuum usually won't kick in for a long time
because the table is so tiny. So some number of old dead row versions is
expected - likewise for your own db.

-- 
Andrew (irc:RhodiumToad)