Re: database-level lockdown - Mailing list pgsql-general

From Greg Sabino Mullane
Subject Re: database-level lockdown
Date
Msg-id 14523323b8cd1c01f3079841e86cff7a@biglumber.com
Whole thread Raw
In response to Re: database-level lockdown  (Filipe Pina <filipe.pina@impactzero.pt>)
List pgsql-general
-----BEGIN PGP SIGNED MESSAGE-----
Hash: RIPEMD160


Filipe Pina said:

> I really can't find any other solution for what I need (in short: make sure
> no transactions are left out due to serialization failures)

I think you may have been too quick to rule out advisory locks as a solution.
Yes, you will need wrappers around all other calls, but extraordinary problems
call for extraordinary solutions.

> I could place the locks from the adapter directly at the outer transaction
> level but I have the feeling that performance would be worse...

Well, performance has really got to take a back seat, given your other
requirements. ;) Locking certainly *could* work - and is arguably the best
solution, as that's what locks are for. Just have your Very Important Transaction
retry, and upon reaching that critical number, exclusively lock just the tables
being used, then try again. If you don't know which tables are being used,
I suggest storing that somewhere your class can find it, or moving away
from such a generic class.

There are other solutions (e.g. forcing conflicting processes to quit
and sleep a second), but the locking one seems the easiest.

- --
Greg Sabino Mullane greg@turnstep.com
End Point Corporation http://www.endpoint.com/
PGP Key: 0x14964AC8 201507050943
http://biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8
-----BEGIN PGP SIGNATURE-----

iEYEAREDAAYFAlWZNeoACgkQvJuQZxSWSshVngCgpzGg7/OXRcyE2JgwDxDTFr9X
o7UAn3ENNgmIVqPpR4j1kyooiu+Ool7A
=6FSv
-----END PGP SIGNATURE-----




pgsql-general by date:

Previous
From: Charles Clavadetscher
Date:
Subject: Re: [pg_hba.conf] publish own Python application using PostgreSQL
Next
From: Dane Foster
Date:
Subject: Why does the range type's upper function behave inconsistently?