Обсуждение: error log, tablespace, wal files

Hello,

     1) what is tablespace in postgres?

     2) one more issue is that, I try to rename or delete some file in /opt/PostgresPlus/9.0/data/base/16428/ directory.  when I restart postgres server, it start normally.  but I thing this is wrong.  I rename or delete file into above directory, that means my database(having 16428 oid) is corrupted.  So How can my postgres server is started normally?? I want to see the log file for that database corruption. Can I see it?? and where???
and please explain me that what's going on in above case??????

     3) I want know about pg_xlog(wal) file creation. That means, size of wal files and how to create that files and time stamp between creation of two different wal files, that means when one file is created on 11:53AM, so how much time  after the second file is created?? can we set that time stamp???

Please any one explain me above points.

Regards,
Karuna karpe.

> Hello,
>
>      1) what is tablespace in postgres?
>
>      2) one more issue is that, I try to rename or delete some file in
> /opt/PostgresPlus/9.0/data/base/16428/ directory.  when I restart postgres
> server, it start normally.  but I thing this is wrong.  I rename or delete
> file into above directory, that means my database(having 16428 oid) is
> corrupted.  So How can my postgres server is started normally?? I want to
> see the log file for that database corruption. Can I see it?? and where???
> and please explain me that what's going on in above case??????
>
>      3) I want know about pg_xlog(wal) file creation. That means, size of
> wal files and how to create that files and time stamp between creation of
> two different wal files, that means when one file is created on 11:53AM, so
> how much time  after the second file is created?? can we set that time
> stamp???

There are people here better qualified to answer this, but I'll take a shot
anyways.

1) tablespaces allow you to physically store tables in different directories
(meaning different disks). So if you have a table that contains gazillions of
records and needs a lot of space, you could add a disk for that, mount it to
some directory on your server and create a tablespace on the disk. Then you'd
tell postgresql to store the large table(s) in that tablespace.

2) I think - and I certainly may be wrong here - that you will encounter the
error only when postgresql tries to access data that should be in those files.
It's kind of like misplacing your car keys - you'll only notice when you need
them.

3) This depends a lot on your settings in postgresql.conf. To my understanding
WAL files are majorly depending on the amount of data written to them and not
based on a time period. If your database does mostly reads, no WAL log will be
needed, thus the files allocated don't fill up quickly. If you have a lot of
write operations, the files will fill quickly and once full the next file (up to
number of WAL files configured) will be created. This one I'm even more fuzzy on
than #2 - so someone here with more knowledge of the internals may set me
right.

On Thu, Nov 10, 2011 at 12:36 AM, Uwe Schroeder <uwe@oss4u.com> wrote:
>
>
>> Hello,
>>
>>      1) what is tablespace in postgres?
>>
>>      2) one more issue is that, I try to rename or delete some file in
>> /opt/PostgresPlus/9.0/data/base/16428/ directory.  when I restart postgres
>> server, it start normally.  but I thing this is wrong.  I rename or delete
>> file into above directory, that means my database(having 16428 oid) is
>> corrupted.  So How can my postgres server is started normally?? I want to
>> see the log file for that database corruption. Can I see it?? and where???
>> and please explain me that what's going on in above case??????

Any idea what was in that?  Have you tried truncating the table that
was behind it?  Assuming it was a table.

On 11/10/2011 02:29 PM, Karuna Karpe wrote:

>       1) what is tablespace in postgres?

It's a way of putting some tables, indexes, etc into a different part of
your system's storage. Sometimes you want to do this to put a really big
table on slower, cheaper storage. Sometimes you might want to put
indexes on really fast but expensive or small and limited storage. It's
also useful for splitting up tables that're often accessed concurrently
so they're on different disk arrays and don't compete with each other
for I/O. There are lots of uses.

>       2) one more issue is that, I try to rename or delete some file in
> /opt/PostgresPlus/9.0/data/base/16428/ directory.

Don't do that!

Do not mess with anything in the PostgreSQL data directory unless you
know *EXACTLY* what you are doing.

>  when I restart
> postgres server, it start normally.  but I thing this is wrong.  I
> rename or delete file into above directory, that means my
> database(having 16428 oid) is corrupted.  So How can my postgres server
> is started normally??

Because it hasn't needed to access that file yet. When it does, it'll
report errors for attempts to access that file but will otherwise
continue to function normally.

PostgreSQL doesn't try to verify the whole database on startup. Imagine
how long that'd take!

> I want to see the log file for that database
> corruption. Can I see it?? and where???
> and please explain me that what's going on in above case??????

You need to figure out what you deleted by looking in pg_catalog, then
attempt to access it.

It'd be nice if there were a built-in tool or utility to verify a
database catalog against the on-disk contents, but it's never supposed
to be necessary. The need only arises when someone does something silly
or when a file system goes bad - and in the latter case, who says it's
the catalog that's correct?

--
Craig Ringer

Thanks to all for giving solution.