[lttng-dev] background information about LTTng timestamps

[Mathieu Desnoyers]

* Sébastien Barthélémy (barthelemy at crans.org) wrote:
> Hello all,

Already fixed this morning by commit:

commit 9080554ff25b08f48762d2ee7cfa3981317e9c1d
Author: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
Date:   Mon Jan 9 11:42:57 2012 -0500

    Fix timestamps for slow-paced event rates
    
    Given that we currently do not implement any timer that provides
    guarantees of 27-bit timestamps overflow detection, force the timestamp
    to 64-bit.
    
    Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>

Please try with LTTng-UST git head.

Thanks,

Mathieu


> 
> I'm using LTTng 2 (only UST for now) to study the behaviour of a realtime
> multiprocess application. Some of the traces I get suggest the synchronization
> between the processes is not working as expected.
> 
> However this might ascome from latencies or even hops in LTTng clock source.
> (Or maybe some bugs in the scripts I use to process the LTTng traces.)
> 
> Before digging deeper in our app synchronization mechanisms, I'd like
> to get a better
> understanding of the way LTTng timestamps my data, how accurate these timestamps
> are and under which conditions I can trust them.
> 
> Is this covered somewhere? I had a look at the source code, but
> without much success.
> Where should I start?
> 
> I added tracepoints with the current time as argument and measured
> that the LTTng
> timestamps where only ~3us behind. Moreover this latency was the same across
> the two processes so this connot account for the ~500us quirks I see.
> 
> Thank you (again) for any help
> 
> Best regards
> -- Sébastien
> 
> _______________________________________________
> lttng-dev mailing list
> lttng-dev at lists.lttng.org
> http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
> 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com