Re: [HACKERS] memory fields from getrusage()

Поиск
Список
Период
Сортировка
От Justin Pryzby
Тема Re: [HACKERS] memory fields from getrusage()
Дата
Msg-id 20170614222840.GB15684@telsasoft.com
обсуждение исходный текст
Ответ на Re: [HACKERS] memory fields from getrusage()  (Robert Haas <robertmhaas@gmail.com>)
Ответы Re: [HACKERS] memory fields from getrusage()  (Robert Haas <robertmhaas@gmail.com>)
Список pgsql-hackers
On Tue, Jun 13, 2017 at 12:16:00PM -0400, Robert Haas wrote:
> It might be worth adding platform-specific code for common platforms.

All I care (which linux happily/happens to support) is maxrss; I was probably
originally interested in this while digging into an issue with hash agg.

I think it's fine to show zeros for unsupported fields; that's what getusage(2)
and time(1) do after all.

pryzbyj@pryzbyj:~$ sh -c 'command time -v find /dev' 2>&1 >/dev/null |grep -Fw 0       User time (seconds): 0.00
Systemtime (seconds): 0.00       Elapsed (wall clock) time (h:mm:ss or m:ss): 0:00.00       Average shared text size
(kbytes):0       Average unshared data size (kbytes): 0       Average stack size (kbytes): 0       Average total size
(kbytes):0       Average resident set size (kbytes): 0       Major (requiring I/O) page faults: 0       Swaps: 0
Filesystem inputs: 0       File system outputs: 0       Socket messages sent: 0       Socket messages received: 0
Signalsdelivered: 0
 

> it would be a good idea to install code specific to Linux that
> displays all and only those values that are meaningful on Linux, and
> (less importantly) similarly for macOS.  Linux is such a common
> platform that reporting bogus zero values and omitting other fields
> that are actually meaningful does not seem like a very good plan.

That has the issue that it varies not just by OS but also by OS version.  For
example PG already shows context switches and FS in/out puts, but they're
nonzero only since linux 2.6 (yes, 2.4 is ancient and unsupported but still).
      ru_nvcsw (since Linux 2.6)      ru_inblock (since Linux 2.6.22)

..and other fields are "currently unused", but maybe supported in the past or
future(?)      ru_ixrss (unmaintained)             This field is currently unused on Linux.

Are you thinking of something like this, maybe hidden away in a separate file
somewhere?

#if defined(__linux__) || defined(BSD)appendStringInfo(&str, "!\t%ld max resident, %ld shared, %ld unshared data, %ld
unsharedstack (kB)\n", r.ru_maxrss, r.ru_ixrss, r.ru_idrss, r.ru_isrss);
 
#elif defined(__darwin__)appendStringInfo(&str, "!\t%ld max resident, %ld shared, %ld unshared data, %ld unshared stack
(kB)\n",r.ru_maxrss/1024, r.ru_ixrss/1024, r.ru_idrss/1024, r.ru_isrss/1024);
 
#endif /* __linux__ */

Justin



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Christian Ullrich
Дата:
Сообщение: Re: [HACKERS] Buildfarm failures on woodlouse (in ecpg-check)
Следующее
От: Petr Jelinek
Дата:
Сообщение: Re: [HACKERS] Get stuck when dropping a subscription duringsynchronizing table