Re: [HACKERS] Rename RECOVERYXLOG to RECOVERYWAL?

2017-09-02 Thread David Steele 
On 9/1/17 7:53 PM, Michael Paquier wrote:
> On Sat, Sep 2, 2017 at 3:06 AM, Robert Haas  wrote:
>> I don't think this really buys us anything.  If we'd applied it to v10
>> maybe, but what do we get out of whacking it around now?
>>
>> "Consistency", I hear you cry!  Fair point.  But we never had a goal
>> of eliminating all internal references to "xlog", just the user-facing
>> ones.  And since RECOVERYXLOG is not documented, I think there's a
>> good argument that it's not user-facing.  You could argue that since
>> it shows up in the file system it's implicitly user-facing, and maybe
>> you're right; if some other committer really wants to make this
>> change, I won't grouse much.  But personally I'd favor leaving it
>> alone to avoid having the behavior change a little bit in every new
>> release.
> 
> I may be wrong, but I would suspect that some backup tools doing
> FS-level backup are checking on the existence of this file and skip
> it. From the point of view of operations, it does not matter much as
> any existing RECOVERYXLOG is unlinked before being replaced by a new
> one, but that would not be nice to add silently 16MB in each backup.

Yes, pgBackRest does have an "offline" mode that can be used (when the
database is shutdown) to do an FS-level backup.

It never occurred to me to exclude RECOVERYXLOG but with 1GB WAL
segments coming in v11 it might be a good idea.

-- 
-David
da...@pgmasters.net


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Re: [HACKERS] Rename RECOVERYXLOG to RECOVERYWAL?

2017-09-01 Thread Michael Paquier 
On Sat, Sep 2, 2017 at 3:06 AM, Robert Haas  wrote:
> I don't think this really buys us anything.  If we'd applied it to v10
> maybe, but what do we get out of whacking it around now?
>
> "Consistency", I hear you cry!  Fair point.  But we never had a goal
> of eliminating all internal references to "xlog", just the user-facing
> ones.  And since RECOVERYXLOG is not documented, I think there's a
> good argument that it's not user-facing.  You could argue that since
> it shows up in the file system it's implicitly user-facing, and maybe
> you're right; if some other committer really wants to make this
> change, I won't grouse much.  But personally I'd favor leaving it
> alone to avoid having the behavior change a little bit in every new
> release.

I may be wrong, but I would suspect that some backup tools doing
FS-level backup are checking on the existence of this file and skip
it. From the point of view of operations, it does not matter much as
any existing RECOVERYXLOG is unlinked before being replaced by a new
one, but that would not be nice to add silently 16MB in each backup.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Re: [HACKERS] Rename RECOVERYXLOG to RECOVERYWAL?

2017-09-01 Thread David Steele 
On 9/1/17 2:06 PM, Robert Haas wrote:
> On Fri, Sep 1, 2017 at 12:57 PM, David Steele  wrote:
>> I searched the various threads on the xlog -> wal rename and I couldn't
>> find any specific mention of why this was not renamed.
>>
>> I have attached a patch in case it was an oversight rather than left
>> as-is on purpose.
> 
> I don't think this really buys us anything.  If we'd applied it to v10
> maybe, but what do we get out of whacking it around now?

I was thinking it would be applied to v10.

> "Consistency", I hear you cry!  Fair point.  But we never had a goal
> of eliminating all internal references to "xlog", just the user-facing
> ones.  And since RECOVERYXLOG is not documented, I think there's a
> good argument that it's not user-facing.  You could argue that since
> it shows up in the file system it's implicitly user-facing, and maybe
> you're right; 

That's exactly my argument, in fact!

> if some other committer really wants to make this
> change, I won't grouse much.  But personally I'd favor leaving it
> alone to avoid having the behavior change a little bit in every new
> release.

Seems like since v10 is still beta and this is not really documented it
wouldn't be that big a deal to make the change.  If nothing else it
might keep the question from coming up in the future.

I'm not going to make a big fuss about it, though.  I noticed it while
testing the v10 support in pgbackRest and thought it was worth bringing up.

Thanks,
-- 
-David
da...@pgmasters.net


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Re: [HACKERS] Rename RECOVERYXLOG to RECOVERYWAL?

2017-09-01 Thread Robert Haas 
On Fri, Sep 1, 2017 at 12:57 PM, David Steele  wrote:
> I searched the various threads on the xlog -> wal rename and I couldn't
> find any specific mention of why this was not renamed.
>
> I have attached a patch in case it was an oversight rather than left
> as-is on purpose.

I don't think this really buys us anything.  If we'd applied it to v10
maybe, but what do we get out of whacking it around now?

"Consistency", I hear you cry!  Fair point.  But we never had a goal
of eliminating all internal references to "xlog", just the user-facing
ones.  And since RECOVERYXLOG is not documented, I think there's a
good argument that it's not user-facing.  You could argue that since
it shows up in the file system it's implicitly user-facing, and maybe
you're right; if some other committer really wants to make this
change, I won't grouse much.  But personally I'd favor leaving it
alone to avoid having the behavior change a little bit in every new
release.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers