[rrd-users] [rrd-developers] RRDTOOL 1.3.6 (rrd_update hangs)

[swami]

Hi,
     To answer your questions , the issue happens with different RRD's each time, no two threads are updating the same RRD, single update is passed to rrd_update_r.

But as i had said in my earlier mail, when system time is changed to be in past , subsequent rrd_update call does something which consumes CPU to 99%. Is it something  normal and if so , what does it do ?



Regards
swami
  ----- Original Message ----- 
  From: swami 
  To: kevin brintnall 
  Cc: rrd-users at lists.oetiker.ch ; rrd-developers at lists.oetiker.ch 
  Sent: Monday, March 29, 2010 6:16 PM
  Subject: Re: [rrd-developers] RRDTOOL 1.3.6 (rrd_update hangs)


  Hi,
        I am happy that i was able to reproduce the issue few minutes ago on seeing your mail .
        I had my double on ntp date/time and I just changed the system time to 1 day ago  explcitily using "date -s" command and my module started hogging CPU to 99% and one of the thread (rrd_update) was doing something.
     
        Is it that when time goes back in the past , rrd_update starts doing something ? If yes, what and how can this be prevented because this is normal when you use NTP server for system time and it could go in the past or future at times.

       Any thoughts / suggestions appreciated .

  Regards
  swami
    ----- Original Message ----- 
    From: kevin brintnall 
    To: swami 
    Cc: rrd-users at lists.oetiker.ch ; rrd-developers at lists.oetiker.ch 
    Sent: Monday, March 29, 2010 5:48 PM
    Subject: Re: [rrd-developers] RRDTOOL 1.3.6 (rrd_update hangs)


    Could you compile with debugging symbols (-g) and re-run the test?

    Are you having this problem consistently with the same RRDs?

    Is it possible that you are updating the same RRD concurrently in two threads ?

    How many updates are you passing to rrd_update_r for a single RRD?

    -kb


    On Mon, Mar 29, 2010 at 7:12 PM, swami <swami at infineta.com> wrote:

      Hi,
           Something weird i am seeing at times with rrdtool version 1.3.6 i am using.
      One of my module starts threads and one of the thread updates an rrd and i find at times its consuming 99% of CPU vs the normal being 2-3 %(2 out of around 15 times i tried restart module).With gdb , i found one of the threads was hung as per the gdb backtrace below

      (gdb) bt
      #0  0x2ca3aa14 in memcpy ()
          at ../ports/sysdeps/unix/sysv/linux/mips/nptl/lowlevellock.h:195
      #1  0x2adcfed0 in rrd_write ()
         from /usr/local/third-party/rrd/lib/librrd_th.so.4
      #2  0x2add4378 in write_to_rras ()
         from /usr/local/third-party/rrd/lib/librrd_th.so.4
      #3  0x2add4ff8 in process_arg ()
         from /usr/local/third-party/rrd/lib/librrd_th.so.4
      #4  0x2add6408 in _rrd_update ()
         from /usr/local/third-party/rrd/lib/librrd_th.so.4
      #5  0x100176e8 in data_collector_service_class_thread_func (ptr=0x0)
          at src/service_class_data_collector.c:832                                      =======================> there is rrd_update_r happening here
      #6  0x2cc0dea0 in start_thread (arg=<value optimized out>)
          at pthread_create.c:263
      #7  0x2ca9639c in __thread_start () from /lib32/libc.so.6
      #8  0x2ca9639c in __thread_start () from /lib32/libc.so.6


      Any idea what this is and any patch for it already available ?


      Regards
      swami

      _______________________________________________
      rrd-developers mailing list
      rrd-developers at lists.oetiker.ch
      https://lists.oetiker.ch/cgi-bin/listinfo/rrd-developers





    -- 
    kevin brintnall =~ /kbrint at rufus.net/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20100330/8d77cfe3/attachment-0001.htm