Re: radosgw high cpu

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Thank you very much.  When I set debug_rgw = 0 from default settings,
std::locale dispeared. And the performance improved a little bit.  I
am using v12.2.4 . Is there any improvment related to log in the later
ceph version?   Or how should I configure it?
And now perf show the following, I think lock cosume too much time,
and cpu utilization is very low. Is there any improvment on this ? :
   4.04%  libtcmalloc_minimal.so.4.4.5  [.] operator new[]
   2.57%  libc-2.17.so                  [.] __strcasecmp_l_avx
   2.15%  libpthread-2.17.so            [.] pthread_mutex_lock
   2.07%  libtcmalloc_minimal.so.4.4.5  [.] tc_deletearray_nothrow
   2.03%  libstdc++.so.6.0.19           [.] 0x000000000005b2ad
   1.87%  libpthread-2.17.so            [.] pthread_mutex_unlock
   1.69%  libstdc++.so.6.0.19           [.] std::basic_string<char,
std::char_traits<char>, std::allocator<char> >::basic_string
   1.61%  libstdc++.so.6.0.19           [.] std::string::assign
   1.60%  libc-2.17.so                  [.] __strlen_sse2_pminub
   1.48%  libc-2.17.so                  [.] __GI___strncasecmp_l
   1.37%  libstdc++.so.6.0.19           [.] __dynamic_cast
   1.31%  libc-2.17.so                  [.] __memcpy_ssse3_back
   1.12%  libc-2.17.so                  [.] __strptime_internal
   1.08%  libc-2.17.so                  [.] vfprintf
   1.06%  libc-2.17.so                  [.] __strftime_internal
   1.02%  libc-2.17.so                  [.] __strncmp_sse42
   0.90%  libc-2.17.so                  [.] __strlen_sse2
   0.88%  libc-2.17.so                  [.] __memcmp_sse4_1
On Tue, Oct 30, 2018 at 11:53 PM kefu chai <tchaikov@xxxxxxxxx> wrote:
>
> On Tue, Oct 30, 2018 at 5:36 PM Will Zhao <zhao6305@xxxxxxxxx> wrote:
> >
> > Hi:
> >     I was trying to figure out how to imporve rgw performance , and I
> > found std::locale comsuming a lot of cpu . Is this normal ? which part
> > of rgw is using this?
>
> Will, probably you could lower the log level and try again? i guess
> that it's our logging subsystem who is creating/destroying
> std::locale. also, what version of Ceph are you testing?
>
> >
> > 18.99%  libstdc++.so.6.0.19   [.] std::use_facet<std::ctype<char> >
> >
> >   18.59%  [kernel]              [k] native_queued_spin_lock_slowpath
> >
> >    8.32%  libstdc++.so.6.0.19   [.] std::locale::locale
> >
> >    5.72%  libstdc++.so.6.0.19   [.] std::locale::~locale
> >
> >    2.49%  libstdc++.so.6.0.19   [.] std::num_put<char,
> > std::ostreambuf_iterator<char, std::char_traits<char> >
> > >::_M_insert_int<long>
> >
> >    2.36%  libstdc++.so.6.0.19   [.] std::has_facet<std::num_put<char,
> > std::ostreambuf_iterator<char, std::char_traits<char> > > >
> >
> >    2.22%  libstdc++.so.6.0.19   [.] std::use_facet<std::num_put<char,
> > std::ostreambuf_iterator<char, std::char_traits<char> > > >
> >
> >    2.00%  libstdc++.so.6.0.19   [.] std::has_facet<std::num_get<char,
> > std::istreambuf_iterator<char, std::char_traits<char> > > >
> >
> >    1.79%  libstdc++.so.6.0.19   [.] std::use_facet<std::num_get<char,
> > std::istreambuf_iterator<char, std::char_traits<char> > > >
> >
> >    1.47%  libstdc++.so.6.0.19   [.] std::num_put<char,
> > std::ostreambuf_iterator<char, std::char_traits<char> >
> > >::_M_insert_int<unsigned long>
>
>
>
> --
> Regards
> Kefu Chai



[Index of Archives]     [CEPH Users]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux