[erlang-bugs] R15B01 erlang:now() jumping ~24 days into the future
Patrik Nyblom
pan@REDACTED
Mon Mar 11 17:26:05 CET 2013
Hi again!
I think I've found it. At least I've found one error, hopefully that's
the one you've also found :)
The sys_gethrtime function has gon new uses in R15 and on, uses where it
is no longer protected by the erts_timeofday_mtx. So - it simply needs
a lock of it's own. This gives a slight performance loss, but that could
be fixed by using GetTickCount64 on win7 and win2008 at least.
Can you try a version of beam.smp.dll with a lock and see if the error
is gone on your machines? If that works, I would also like you to try an
optimized version, but let's first make sure we have the bug nailed down :)
In my dropbox, there's a beam.smp.dll. If you replace
$ERL_ROOT/erts-5.10.1/bin/beam.smp.dll with that one and then start
werl, the slogan should contain [source-be0da3e]. It is for 64bit
windows. The public dropbox URL is:
http://dl.dropbox.com/u/17212223/beam.smp.dll
This should work without any special messages or such, giving a working
erlang:now/0. If it starts sending strange ERROR REPORT's about ticks
moving slightly backwards, we have a more complicated bug, but I haven't
seen any such messages since i added proper locking.
If it's possible for you to test this, I would be immensely grateful!
Cheers,
/Patrik
On 03/07/2013 04:37 PM, Patrik Nyblom wrote:
> Hi Garret!
>
> I've been able to reproduce it on my freshly installed Win2008
> machine! Great, now I only need to debug it and find the error :)
>
> I'll get back to you as soon as I feel I have a fix - it might take a
> few days, given the relatively long turn around time, but we'll get there!
>
> Thank you for all the help and information!
>
> Cheers,
> /Patrik
>
> On 03/05/2013 09:10 PM, Garret Smith wrote:
>> On the same machine with the same steps as previous, I reproduced the
>> time jump on R16B.
>> This time the jump happened with a <5 sec delta btw now() and
>> os:timestamp().
>> Still jumping ~2126000 seconds.
>>
>> -Garret
>>
>>
>> On Tue, Mar 5, 2013 at 11:20 AM, Garret Smith <garret.smith@REDACTED
>> <mailto:garret.smith@REDACTED>> wrote:
>>
>> The gist https://gist.github.com/garret-smith/5087169 is updated
>> with a slightly better version. I was able to reproduce the jump
>> in less than an hour. I also did some more things to perturb the
>> timing code while the test program was running.
>>
>> Here is the latest info, everything I can think of that may have
>> the slightest effect:
>> * R15B01 64-bit build
>> * Pacific time zone (GMT -8)
>> * Xeon E5405 in an HP DL160
>> * no arguments to erl.exe
>> * bursty, high CPU load, >75% memory used by other software
>> * running Observer on the test VM displaying the "Load Charts" tab
>> * made some small adjustments (~ 60 seconds) to the system clock
>> while running the tests - now() and os:timestamp() behaved as
>> expected, initially showing a delta and slowly converging
>> * w32tm /resync to fix the system clock some time after
>> perturbing it
>>
>> The time jump in now() occurred when now() was ~9 seconds behind
>> os:timestamp() as reported by the new test program.
>>
>> I'm starting to look at R16B now.
>>
>> -Garret Smith
>>
>>
>> On Tue, Mar 5, 2013 at 8:37 AM, Garret Smith
>> <garret.smith@REDACTED <mailto:garret.smith@REDACTED>> wrote:
>>
>> I haven't seen anything unexpected in os:timestamp(). No
>> jumps at all.
>>
>> CPU is an Intel Xeon X3430.
>>
>> I have reproduced it in the LosAngeles/Pacific Time (GMT -8)
>> and US East coast time zone (GMT -5).
>>
>> I have not yet tried R16B. I'll be starting that today. I'm
>> also trying to improve the test program, since it's taking
>> quite a long time between jumps for me as well. I'll let you
>> know as soon as I have a better one.
>>
>> You have no idea how relieved I am that you are looking into
>> this!
>>
>> Thanks,
>> Garret Smith
>>
>>
>> On Tue, Mar 5, 2013 at 3:06 AM, Patrik Nyblom <pan@REDACTED
>> <mailto:pan@REDACTED>> wrote:
>>
>> Hi again...
>>
>> I'm not sure about one thing. What happens to
>> os:timestamp() during these jumps? Does it stay on track
>> or does it also jump around?
>>
>> I've tried to reproduce it with your program, but has not
>> yet succeeded. Have you seen this on the R16B release as
>> well?
>>
>> Is the hardware in any way fancy (like a lot of cores,
>> some new processor I don't have or something else?) or is
>> there anything else special about the machine? Also the
>> time zone you're running in would be interesting, as
>> there is some time zone specific code there...
>>
>> I would really like to be able to reproduce it so you
>> don't have to do all the tests at your site, it might end
>> up being really time consuming for you if I make to many
>> mistakes :)
>>
>> Cheers,
>> /Patrik
>>
>>
>>
>> On 03/05/2013 08:50 AM, Patrik Nyblom wrote:
>>> Hi!
>>>
>>> On 03/05/2013 02:26 AM, Garret Smith wrote:
>>>> I have been beating my head against a wall for weeks
>>>> tracking down spooky behaviour[sic] in one of our
>>>> production systems. I finally tracked it down to
>>>> "jumps" in the times returned by erlang:now(), causing
>>>> all timers in the system to expire at once. I have
>>>> witnessed this bug on R15B01, both 64 and 32-bit
>>>> versions running on Windows Server 2008 R2, both on
>>>> bare metal and VirtualBox VM.
>>>>
>>>> The time jump is always around 2126000 seconds, or a
>>>> little over 24 days. The now() time does not try to
>>>> converge with os:timestamp() as the documentation
>>>> suggests, and as I confirmed it does if you just change
>>>> the system clock.
>>>>
>>>> Another VM running concurrently on the same machine but
>>>> with little load (diagnostic node & production node)
>>>> did not time jump.
>>>>
>>>> Higher load seems to make the time jumps happen more often.
>>>>
>>>> Frequency between time jumps varies between seconds and
>>>> hours, but when a jump occurs, it is always 2126000 +
>>>> (9 to 26) seconds.
>>>>
>>>> I never see the jump in logfile timestamps that use
>>>> os:timestamp() for tagging log messages. I had to
>>>> start tracing a production node before I caught the
>>>> jump. Here are some lines from a trace, where the
>>>> timestamp in trace_ts is printed using
>>>> calendar:now_to_local_time() and then in raw tuple format:
>>>>
>>>> 2013-4-16 21:40:1.993399|{1366,173601,993399}
>>>> 2013-4-16 21:40:1.993400|{1366,173601,993400}
>>>> 2013-5-11 12:13:41.986961|{1368,299621,986961}
>>>> 2013-5-11 12:13:41.986962|{1368,299621,986962}
>>>>
>>>> then a bit later...
>>>>
>>>> 2013-5-11 12:36:19.955129|{1368,300979,955129}
>>>> 2013-5-11 12:36:19.955130|{1368,300979,955130}
>>>> 2013-6-5 3:9:49.538830|{1370,426989,538830}
>>>> 2013-6-5 3:9:49.538833|{1370,426989,538833}
>>>>
>>> Gah! That's obviously not supposed to happen...
>>>> I captured many such jumps over the course of a day or
>>>> so. Obviously from the dates, 2 jumps happened before
>>>> I started tracing.
>>>>
>>>> I was able to reproduce the bug, though not as
>>>> efficiently as my production system, with the following
>>>> sample program:
>>>> https://gist.github.com/garret-smith/5087169
>>>>
>>>> It took over an hour of runtime before the first time
>>>> jump. I am working on a better way to reproduce it at
>>>> the moment, but it's hard to test the test with a bug
>>>> so intermittent.
>>>>
>>>> I am also testing various other VM versions. My first
>>>> hope was that this was limited to the 64-bit version
>>>> where we first encountered the problem, but a change to
>>>> the 32-bit version has only made the problem happen
>>>> less often, not eliminated it.
>>>>
>>>> We never saw this bug with R14B03 which we were running
>>>> previously to R15B01. However, system load is
>>>> different so I can't make a direct comparison. I did
>>>> notice a few significant updates to the Windows time
>>>> related code between R14B03 and R15:
>>>>
>>>> git log sys_time.c
>>>>
>>>> commit 46eb4359b05b220861453a869dc734480ec045a6
>>>> Author: Patrik Nyblom <pan@REDACTED
>>>> <mailto:pan@REDACTED>>
>>>> Date: Tue Dec 6 19:07:16 2011 +0100
>>>>
>>>> Emulate localtime, gmtime and mktime to enable
>>>> negative time_t
>>>>
>>>> commit 913f05af100e98a8665bbb6168e89fbcfe4ece75
>>>> Author: Bj<C3><B6>rn-Egil Dahlberg <egil@REDACTED
>>>> <mailto:egil@REDACTED>>
>>>> Date: Fri Dec 2 15:25:06 2011 +0100
>>>>
>>>> Teach windows sys_localtime_r
>>>>
>>>>
>>> Yep, that's me... But even if I gave a totally weird
>>> time back from those, the erlang:now logic should have
>>> stopped this from happening. I'll try to reproduce using
>>> your example program. If nothing else helps, I'll
>>> instrument a VM that gives som traces in the time code...
>>>> I am completely stumped. What can I do next to help
>>>> track down the source of the bug?
>>>>
>>> Unfortunately, so am I. Especially weird that it's load
>>> related... Maybe something is not locked as it should be...
>>>> Thanks,
>>>> Garret Smith
>>> Thanks for reporting, I'll get back to you!
>>>
>>> Cheers,
>>> /Patrik
>>>>
>>>>
>>>> _______________________________________________
>>>> erlang-bugs mailing list
>>>> erlang-bugs@REDACTED <mailto:erlang-bugs@REDACTED>
>>>> http://erlang.org/mailman/listinfo/erlang-bugs
>>>
>>>
>>>
>>> _______________________________________________
>>> erlang-bugs mailing list
>>> erlang-bugs@REDACTED <mailto:erlang-bugs@REDACTED>
>>> http://erlang.org/mailman/listinfo/erlang-bugs
>>
>>
>> _______________________________________________
>> erlang-bugs mailing list
>> erlang-bugs@REDACTED <mailto:erlang-bugs@REDACTED>
>> http://erlang.org/mailman/listinfo/erlang-bugs
>>
>>
>>
>>
>
>
>
> _______________________________________________
> erlang-bugs mailing list
> erlang-bugs@REDACTED
> http://erlang.org/mailman/listinfo/erlang-bugs
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-bugs/attachments/20130311/9c46cc6d/attachment.htm>
More information about the erlang-bugs
mailing list