[ltt-dev] Events lost

Matthew Khouzam matthew.khouzam at ericsson.com
Mon Jan 17 11:44:32 EST 2011

On 11-01-17 11:17 AM, Alessandro Morari wrote:
> Hi all,
> I'm using LTTng ( kernel and LTTV ( 0.12.20) to obtain 
> traces of an application running in a Blue Gene/P system.
> This is a run with 4 cores for each node. When running "LTTV -m 
> textDump ..." to process the traces I get the following message:
> ** (process:1002): WARNING **: 1842089 events lost so far in tracefile 
> trace116/net_2 at block 90
> Do you know how can I avoid this events to be lost and this message to 
> appear?

The message is very important, you don't want to lose it.

As for not getting it, you probably busted your buffer size.
trace116/net_2 <- this means that net of cpu3 had a buffer overrun.

The way I avoid that is by giving each channel more buffers.
  lttctl -C -o channel.net.bufsize=$((1<<23))  something like this 
should help a lot.

> Thank you,
> Alessandro
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.casi.polymtl.ca/pipermail/lttng-dev/attachments/20110117/de806f5e/attachment-0003.htm>

More information about the lttng-dev mailing list