Re: initdb and fsync - Mailing list pgsql-hackers

From Cédric Villemain
Subject Re: initdb and fsync
Date
Msg-id 201203171748.38738.cedric@2ndquadrant.com
Whole thread Raw
In response to Re: initdb and fsync  (Andres Freund <andres@anarazel.de>)
Responses Re: initdb and fsync  (Jeff Davis <pgsql@j-davis.com>)
List pgsql-hackers
<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">Levendredi 16 mars 2012 16:51:04, Andres Freund a écrit :<p style=" margin-top:0px;
margin-bottom:0px;margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> On
Friday,March 16, 2012 04:47:06 PM Robert Haas wrote:<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > On Fri, Mar 16, 2012 at 6:25 AM,
AndresFreund <andres@anarazel.de> wrote:<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > >> > How are the results
withsync_file_range(fd, 0, 0,<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> > >> > SYNC_FILE_RANGE_WRITE)?<p style="
margin-top:0px;margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>> >> <p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> > >> That is much faster than using fadvise. It
goesdown to ~2s.<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0;
text-indent:0px;-qt-user-state:0;">> > >> <p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > >> Unfortunately, that's
non-portable.Any other ideas? 6.5s a little on<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > >> the annoying side (and
causessome disconcerting sounds to come from<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > >> my disk), especially when
we_know_ it can be done in 2s.<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> > > <p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > > Its not like
posix_fadviseis actually portable. So I personally don't<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > > see a problem with that,
but...<pstyle=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0;
text-indent:0px;-qt-user-state:0;">> > <p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> > Well, sync_file_range only works on
Linux,and will probably never<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> > work anywhere else. posix_fadvise() at least has a
chanceof being<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0;
text-indent:0px;-qt-user-state:0;">> > supported on other platforms, being a standard and all that. Though I<p
style="margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>> see that my Mac has neither. :-(<p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> <p style="
margin-top:0px;margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>I would suggest adding a wrapper function like:<p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">>
pg_hint_writeback_flush(fd,off, len);<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> <p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> which then is something
like<pstyle=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0;
text-indent:0px;-qt-user-state:0;">> <p style=" margin-top:0px; margin-bottom:0px; margin-left:0px;
margin-right:0px;-qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> #if HAVE_SYNC_FILE_RANGE<p style="
margin-top:0px;margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>sync_file_range(fd, off, len, SYNC_FILE_RANGE_WRITE);<p style=" margin-top:0px;
margin-bottom:0px;margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">>
#elseifHAVE_POSIX_FADVISE<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> posix_fadvise(fd, off, len, POSIX_FADV_DONTNEED);<p style="
margin-top:0px;margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>#else<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">> #endif<p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">> <p style="
margin-top:0px;margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">>To my knowledge posix_fadvise currently is only supported on linux btw...<p
style="-qt-paragraph-type:empty;margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;"><br /><p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">I agree with Andres.<p
style="-qt-paragraph-type:empty;margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;"><br /><p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">I believe we should use
sync_file_range(_before?) with linux.<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">And we can use posix_fadvise_dontneed on other kernels.<p
style="-qt-paragraph-type:empty;margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;"><br /><p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">FADVISE_DONTNEED does start a
writeback(it may decide not to do it too), but the primer objective of posix_fadvise_dontneed is not to make sync()
faster.We just have writeback and sync() calls challenged together and we can face situation where linux does not
handlethat so well. (depends on linux 2.6.18 or 32 or 3.2 or ...)<p style="-qt-paragraph-type:empty; margin-top:0px;
margin-bottom:0px;margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;"><br /><p
style="-qt-paragraph-type:empty;margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;"><br /><p style=" margin-top:0px; margin-bottom:0px;
margin-left:0px;margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">-- <p style=" margin-top:0px;
margin-bottom:0px;margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;">Cédric
Villemain+33 (0)6 20 30 22 52<p style=" margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px;
-qt-block-indent:0;text-indent:0px; -qt-user-state:0;">http://2ndQuadrant.fr/<p style=" margin-top:0px;
margin-bottom:0px;margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px;
-qt-user-state:0;">PostgreSQL:Support 24x7 - Développement, Expertise et Formation 

pgsql-hackers by date:

Previous
From: Andrew Dunstan
Date:
Subject: Re: Regarding column reordering project for GSoc 2012
Next
From: "Atri Sharma"
Date:
Subject: Re: Regarding column reordering project for GSoc 2012