Обсуждение: pg_ugprade use of --check and --link

Поиск
Список
Период
Сортировка

pg_ugprade use of --check and --link

От
Bruce Momjian
Дата:
In a private bug report, I have realized that if you are eventually
going to be using link mode with pg_upgrade, and you run --check mode,
you should use --link with --check to check that both clusters are on
the same file system.

I have documented this with the attached, applied patch, and backpatched
it to 9.2.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + It's impossible for everything to be true. +



Re: pg_ugprade use of --check and --link

От
Alvaro Herrera
Дата:
Bruce Momjian wrote:
> In a private bug report, I have realized that if you are eventually
> going to be using link mode with pg_upgrade, and you run --check mode,
> you should use --link with --check to check that both clusters are on
> the same file system.

Would it make sense to run the filesystem check anyway, and if the
check fails say something like "this will fail in --link mode, although
copy mode might work" ?

-- 
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services



Re: pg_ugprade use of --check and --link

От
Bruce Momjian
Дата:
On Thu, Jun  6, 2013 at 10:24:30AM -0400, Alvaro Herrera wrote:
> Bruce Momjian wrote:
> > In a private bug report, I have realized that if you are eventually
> > going to be using link mode with pg_upgrade, and you run --check mode,
> > you should use --link with --check to check that both clusters are on
> > the same file system.
> 
> Would it make sense to run the filesystem check anyway, and if the
> check fails say something like "this will fail in --link mode, although
> copy mode might work" ?

I thought about that, but the problem is that the return value of
--check should be success/fail.  We could do the check always, and issue
a warning, but you would still need to use --link with --check if you
wanted a valid return value --- that seemed just too awkward. 
Documenting the --link requirement seemed clearer.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + It's impossible for everything to be true. +