Postgres Pro
Downloads
Home   >   mailing lists

Thread: Two large files instead of one?

Two large files instead of one?

From
Patrik Kudo
Date:
Hi!

I have a database in production with a lot of large objects happening. A
while ago we noticed that the database directory contained two files
with the name of the pg_largeobject oid, of which one had the sufix
".1". Is this a valid file? Is the table paged when reaching a certain
size or is it some kind of lost transaction/sort file? I noticed that
the timestamp of the file both are from today so I guess they should
both be there... I just want to make sure =)

More info:

select version();
                                version
---------------------------------------------------------------------
  PostgreSQL 7.2.1 on i386-portbld-freebsd4.4, compiled by GCC 2.95.3


  select relname from pg_class where oid = 16404;
     relname
----------------
  pg_largeobject

# ls -l|sort +4 -n |tail -n2
-rw-------  1 pgsql  pgsql   465461248 Aug 28 09:20 16404.1
-rw-------  1 pgsql  pgsql  1073741824 Aug 28 01:08 16404


Any info appreciated.

Best regards,
Patrik Kudo

Re: Two large files instead of one?

From
dima
Date:
postgres creates a new file (called .1, .2 etc) when the size of the
previous one gets larger than 1GB (as you can see in your ls output)

> I have a database in production with a lot of large objects happening. A
> while ago we noticed that the database directory contained two files
> with the name of the pg_largeobject oid, of which one had the sufix
> ".1". Is this a valid file? Is the table paged when reaching a certain
> size or is it some kind of lost transaction/sort file? I noticed that
> the timestamp of the file both are from today so I guess they should
> both be there... I just want to make sure =)
>
> More info:
>
> select version();
>                                version
> ---------------------------------------------------------------------
>  PostgreSQL 7.2.1 on i386-portbld-freebsd4.4, compiled by GCC 2.95.3
>
>
>  select relname from pg_class where oid = 16404;
>     relname
> ----------------
>  pg_largeobject
>
> # ls -l|sort +4 -n |tail -n2
> -rw-------  1 pgsql  pgsql   465461248 Aug 28 09:20 16404.1
> -rw-------  1 pgsql  pgsql  1073741824 Aug 28 01:08 16404

Re: Two large files instead of one?

From
Patrik Kudo
Date:
Ok, that's what I suspected. Now I know. Thanks

Regards,
Patrik Kudo

dima wrote:
> postgres creates a new file (called .1, .2 etc) when the size of the
> previous one gets larger than 1GB (as you can see in your ls output)
>
>> I have a database in production with a lot of large objects happening.
>> A while ago we noticed that the database directory contained two files
>> with the name of the pg_largeobject oid, of which one had the sufix
>> ".1". Is this a valid file? Is the table paged when reaching a certain
>> size or is it some kind of lost transaction/sort file? I noticed that
>> the timestamp of the file both are from today so I guess they should
>> both be there... I just want to make sure =)
>>
>> More info:
>>
>> select version();
>>                                version
>> ---------------------------------------------------------------------
>>  PostgreSQL 7.2.1 on i386-portbld-freebsd4.4, compiled by GCC 2.95.3
>>
>>
>>  select relname from pg_class where oid = 16404;
>>     relname
>> ----------------
>>  pg_largeobject
>>
>> # ls -l|sort +4 -n |tail -n2
>> -rw-------  1 pgsql  pgsql   465461248 Aug 28 09:20 16404.1
>> -rw-------  1 pgsql  pgsql  1073741824 Aug 28 01:08 16404
>
>
>
>