[lttng-dev] Restriction on channel sub-buffer size?

[Jérémie Galarneau]

On Sun, Mar 20, 2016 at 3:52 PM, Sean Heelan <seanheelan at gmail.com> wrote:
> Hi all,
>
> I recently encountered a situation whereby tracing in a tight loop was
> generating a very large number of events, thus resulting in quite a few
> discarded events. To resolve this I attempted to create a new channel with a
> larger buffer size as follows:
>
> lttng create cc_session -o $trace_output_path
> lttng enable-channel -u custom-channel--subbuf-size 4M
> lttng enable-event -u -c custom-channel cc_provider:*
> lttng start
>
> /run/target
>
> lttng stop
> lttng destroy
>
> After running the above the trace is empty. However, if I change the size to
> 1M or 2M it works fine. Is there a restriction on the sub-buffer size? I
> also notice a similar behaviour if I increase the number of subbuffers above
> 8 while using a 1M buffer size.
>
> I'm still losing around 80k events at a sub-buffer size of 2M (down from 1.2
> million at the default buffer size buffer size of 4096) so ideally I'd like
> to use a larger buffer size (assuming it's possible).

Hi Sean,

There are no such restrictions on buffer sizes. I have tried your
example here using one of the examples in the lttng-ust tree and I
can't reproduce the problem.

Could you try reproducing with the session daemon launched in verbose
consumer mode?
$ lttng-sessiond -vvv --verbose-consumer

Also, can you launch the application with the environment variable
LTTNG_UST_DEBUG=1.

I'll have a look at the logs.

Thanks!
Jérémie

>
> Some potentially relevant details:
> * I'm running the target inside docker, version 1.10.3
> * The container is Ubuntu 14.04
> * I'm using LTTng from git HEAD
>
> If I can provide any other useful information let me know!
>
> Thanks,
> Sean
>
> _______________________________________________
> lttng-dev mailing list
> lttng-dev at lists.lttng.org
> https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
>



-- 
Jérémie Galarneau
EfficiOS Inc.
http://www.efficios.com