On 27 Mar 2012, at 5:53pm, Pete wrote:
> Interesting. Does that mean any open transaction other than the VACUUM
> transaction? I'm still confused.
This is my fault. I forgot VACUUM was an exception. Please ignore what I
wrote and believe what Peter wrote: You must
t 9:00 AM, <sqlite-users-requ...@sqlite.org> wrote:
> > Date: Mon, 26 Mar 2012 10:25:49 -0700 (PDT)
> > From: Peter Aronson <pbaron...@att.net>
> > To: General Discussion of SQLite Database <sqlite-users@sqlite.org>
> > Subject: Re: [sqlite] VACUUMing large DBs
&g
;pbaron...@att.net>
> To: General Discussion of SQLite Database <sqlite-users@sqlite.org>
> Subject: Re: [sqlite] VACUUMing large DBs
> Message-ID:
><1332782749.22198.yahoomai...@web180307.mail.gq1.yahoo.com>
> Content-Type: text/plain; charset=iso-8859-1
>
> Actual
m: Pete <p...@mollysrevenge.com>
To: sqlite-users@sqlite.org
Sent: Mon, March 26, 2012 10:14:32 AM
Subject: Re: [sqlite] VACUUMing large DBs
SHould a VACUUM command be wrapped in a transaction, or is that done
automatically?
--
Pete
___
sqlite-users mai
On 26 Mar 2012, at 6:14pm, Pete wrote:
> SHould a VACUUM command be wrapped in a transaction, or is that done
> automatically?
All SQLite commands, even ones like SELECT which don't change anything, are
actually executed inside a transaction. If you've already opened
SHould a VACUUM command be wrapped in a transaction, or is that done
automatically?
--
Pete
___
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
Very nice! Thanks !
But then - can you turn journaling off and then run a VACUUM and have it
run as a 2-step instead of a 3-step?
On Thu, Mar 22, 2012 at 3:25 PM, Petite Abeille wrote:
>
> On Mar 22, 2012, at 11:19 PM, Udi Karni wrote:
>
> > Is there a way to run
On Mar 22, 2012, at 11:19 PM, Udi Karni wrote:
> Is there a way to run NOLOGGING in SQlite syntax - which means that if
> something in the destination table/DB fails - you are prepared to just drop
> it and start over?
PRAGMA journal_mode=off
http://sqlite.org/pragma.html#pragma_journal_mode
For the time being - I have been avoiding the VACUUM of very large DBs by
creating a new iteration of the table/DB for each transformation instead of
using UPDATE/DELETE (given that I only have 1 table per DB) -
(1) create new DB_V2 / Table_V2
(2) attach DB_V1 / Table_V1
(3) insert into Table_V2
On Tue, Mar 20, 2012 at 8:25 PM, Jay A. Kreibich wrote:
> On Tue, Mar 20, 2012 at 01:59:59PM -0700, Udi Karni scratched on the wall:
>> Is there a way to go directory from "original" to "journal/final" -
>> skipping the creation of the Temp version?
>
> No, it requires all three
On 21 Mar 2012, at 3:08am, Udi Karni wrote:
> The 240GB SSD drives are pretty reasonably priced and would suffice for
> most tables. I'm just wondering how long before Flash Write Fatigue sets in
> and you need a replacement.
Nobody knows yet. We've done a ton of tests
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 20/03/12 20:08, Udi Karni wrote:
> Thanks! I got one and tried - and it seems to improve overall
> performance about 2X. Very cool.
Depending on your backups and tolerance for data loss, you can also do
things like RAID 0 striping across
On Tue, Mar 20, 2012 at 01:59:59PM -0700, Udi Karni scratched on the wall:
> Hello,
>
> I am creating large DBs - each with a single table (magnitude of a few
> hundred million rows / 100GB). It takes a few transformations to get to the
> final product. When done - I VACUUM the final result.
>
>
Thanks! I got one and tried - and it seems to improve overall performance
about 2X. Very cool.
The 240GB SSD drives are pretty reasonably priced and would suffice for
most tables. I'm just wondering how long before Flash Write Fatigue sets in
and you need a replacement.
On Tue, Mar 20,
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 20/03/12 13:59, Udi Karni wrote:
> And a more general question. My PC has 8GB of RAM. I am considering
> getting a much larger machine that can take upwards of 100-200GB of
> RAM.
I'd recommend getting one or more SSDs instead (also a lot
15 matches
Mail list logo