Re: contrib/start-scripts/linux failing on RHEL 6 with ~9.3 because of missing oom_score_adj

Michael Paquier <michael(dot)paquier(at)gmail(dot)com> writes:
> On Sun, Apr 26, 2015 at 8:47 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> This patch seems completely wrong, as it is forcing use of the oom
>> adjustment, whereas the intention is that the user should choose
>> if they want to use it; cf /contrib/start-scripts/linux lines 43-55.

> On REL9_3_STABLE, the condition the startup script has is that:
> test x"$OOM_SCORE_ADJ" != x && echo "$OOM_SCORE_ADJ" > /proc/self/oom_score_adj
> test x"$OOM_ADJ" != x && echo "$OOM_ADJ" > /proc/self/oom_adj

Yeah, but by default $OOM_SCORE_ADJ is unset so that the echo is not
executed. What you are proposing is that if the user has accidentally
uncommented the wrong one of the two variables, the script should silently
do nothing instead of failing. I think that's an actively bad idea,
because then it would be far too hard to notice that you'd misconfigured
it.

In any case, I think changing the behavior of the script in stable
branches would be unwise. People generally will use these things as
templates, they won't just blindly copy them (at least I hope not);
and they're unlikely to overwrite an existing live script file when
installing a minor update release.

We have changed the script, hopefully for the better, in HEAD --- if
you think the behavior still leaves something to be desired, debating
that would certainly be fair game. (I notice that HEAD *is* using a
"test -e", which according to my argument above maybe isn't a good thing.)

regards, tom lane