On Wed, Oct 30, 2019 at 6:00 PM rihad <rihad@mail.ru> wrote:
> Hi, all. Why is it normally suggested to stop the server, upgrade it,
> then start it? Wouldn't it be easier & quicker to simply upgrade the
> package in-place and restart the service? On OSen that allow
> modification of currently running binaries, which is most Unix OS, M$
> Windows being a notable exception )
Not sure exactly why, and whether postmaster protects itself from
this, but in many years of using linux, which supports it, I've
encountered my share of (minor) problems when doing that, typically
due to the program not having loaded all the modules it can use (
firefox is a classic for me, as I have it normally open while I do my
background updates, nothing a restart does not solve, but I would get
nervous doing that to the database ).
That being said, I've upgraded my TEST servers without stopping them (
although lately the upgrade scripts do a restart at the end ), but for
production I prefer to download everything and prepare as much as I
can and do a stop-finish upgrade-start, the last ( file extracting )
phase is normally much faster than the server restart machinery, so no
much is gained by doing it in paralell.
And other thing, MODIFICATION of currently running binaries is BAD,
and IIRC many OS use them for paging and will not be happy ( maybe in
these days they transparently switch to normal swap ). What you
normally want is unlinking and replacing, modification of the names
aka dir contents (not being able to do this is one of my strongest
dislikes of windows, it makes so many things so much simpler ).
Francisco Olarte.