From mboxrd@z Thu Jan  1 00:00:00 1970
From: compudj@krystal.dyndns.org (Mathieu Desnoyers)
Date: Tue, 1 Nov 2011 14:41:02 -0400
Subject: [ltt-dev] Flight-recording mode
In-Reply-To: <4EAE74A4.2000800@keymile.com>
References: <4EA9A8ED.2010303@keymile.com> <20111028073928.GA4855@Krystal>
	<4EAA65CB.5010009@keymile.com> <20111028082448.GA11499@Krystal>
	<4EAE74A4.2000800@keymile.com>
Message-ID: <20111101184102.GA19895@Krystal>

* Gerlando Falauto (gerlando.falauto at keymile.com) wrote:
> On 10/28/2011 10:24 AM, Mathieu Desnoyers wrote:
>> * Gerlando Falauto (gerlando.falauto at keymile.com) wrote:
>>> On 10/28/2011 09:39 AM, Mathieu Desnoyers wrote:
>>>> * Gerlando Falauto (gerlando.falauto at keymile.com) wrote:
>>> ...
>>>>> How can I make sure that only the latest N seconds of traces will be
>>>>> saved? The answer should be flight-recording mode, but how do I really
>>>>> do it?
>>>>
>>>> I discussed this with David this morning, and we'll need to enhance
>>>> ltt-sessiond in lttng-tools to ensure it does not send the streams to
>>>> the consumer daemon upon trace start for overwrite mode channels, but
>>>> just on trace stop. This is lacking currently.
>>>
>>> OK, thanks!
>>> Does that also apply to LTTng 0.x (since I got the same impression)?
>>>
>>> For 0.x I came up with the idea of killing lttd but I don't know if
>>> a) it's a good idea;
>>> b) it would do the trick
>>
>> for 0.x it should all just work (TM) ;)
>
> Yeah, it sorta-kinda-most-of-the-times-almost works (R). :-)

:)

>
> I am having problems seeing those traces on LTTv though.
> I had to come up with a workaround by patching LLTv and doing some weird  
> stuff after loading the trace in order to have LTTv not crash.

That's not supposed to be necessary. Which lttv version are you using ?
A gdb backtrace of the crash, with lttv compiled with make clean; make
CFLAGS=-g   would be helpful.

>
> And also, half the times I cannot see some of the process names (of  
> course, the ones I am really interested in). Is there any explanation  
> for this?

The flight recording mode does not have the events of the statedump done
at the beginning of the trace. You might wait to try saving the
statedump-related events into separate channels, which are in "normal"
mode (not flight recorder).

>
>> by specifying flight recorder mode, lttctl is supposed to launch lttd
>> only on non-overwrite mode buffers (metadata) first, and only upon trace
>> teardown does it hook lttd on the flight recorder buffers.
>
> If you don't mind me asking, what's the point of having lttd running in  
> background in that case (all channels in flight recorder mode)?

the metadata channel is always in "normal" mode (not flight recorder),
so we need a consumer to get the information as the trace is recorded.

Best regards,

Mathieu

>
> Thanks!
> Gerlando
>

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