[lttng-dev] [barectf] How to ensure that last events before a crash get recorded

[lttng-dev] [barectf] How to ensure that last events before a crash get recorded

[RADERMACHER Ansgar via lttng-dev]

[-- Attachment #1.1: Type: text/plain, Size: 649 bytes --]

Hi,

when doing tracing with barectf, the trace elements are written into a buffer first and only written when the buffer is full - or if the function barectf_platform_fs_fini gets called.
In case of a crash, it's therefore possible to loose some events. What is the best option to prevent this issue? I've used signal handlers that call the function barectf_platform_fs_fini. This seems to work well with Linux, but is not portable. When using the provided sample platform, there is also the option to reduce the buffer size, but this is not ideal, as trace events that are too big for the buffer are not written.

Best regards

Ansgar



[-- Attachment #1.2: Type: text/html, Size: 1094 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

Re: [lttng-dev] [barectf] How to ensure that last events before a crash get recorded

[Philippe Proulx via lttng-dev]

[-- Attachment #1.1: Type: text/plain, Size: 1480 bytes --]

On Thu, Mar 4, 2021 at 10:28 RADERMACHER Ansgar via lttng-dev <
lttng-dev@lists.lttng.org> wrote:

> Hi,
>
> when doing tracing with barectf, the trace elements are written into a
> buffer first and only written when the buffer is full - or if the function
> barectf_platform_fs_fini gets called.
> In case of a crash, it's therefore possible to loose some events. What is
> the best option to prevent this issue? I've used signal handlers that call
> the function barectf_platform_fs_fini. This seems to work well with Linux,
> but is not portable. When using the provided sample platform, there is also
> the option to reduce the buffer size, but this is not ideal, as trace
> events that are too big for the buffer are not written.
>

First question: why are you using barectf on Linux vs LTTng?

LTTng supports writing the sub-buffers to NVRAM and, if the system crashes,
convert those sub-buffers to a CTF trace with the lttng-crash utility. See <
https://lttng.org/docs/v2.12/#doc-persistent-memory-file-systems>.

With NVRAM, you could do this with barectf with a corresponding platform
using it. It's always up to the platform author to decide where buffers are
and where complete packets go; it's not a feature of barectf as such.

Hope it helps,

Phil


> Best regards
>
> Ansgar
>
>
> _______________________________________________
> lttng-dev mailing list
> lttng-dev@lists.lttng.org
> https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
>
-- 
Philippe Proulx

[-- Attachment #1.2: Type: text/html, Size: 3234 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

Re: [lttng-dev] [barectf] How to ensure that last events before a crash get recorded

[RADERMACHER Ansgar via lttng-dev]

[-- Attachment #1.1: Type: text/plain, Size: 2007 bytes --]

Hi Philippe,

thanks for your response. The reason not to use LTTng is simple: while I develop on Linux, our client is chiefly using Windows with the mingw environment.

Best regards

Ansgar

________________________________
From: P
hilippe Proulx [eeppeliteloop@gmail.com]
Sent: Friday, March 05, 2021 13:11
To: RADERMACHER Ansgar
Cc: lttng-dev@lists.lttng.org
Subject: Re: [lttng-dev] [barectf] How to ensure that last events before a crash get recorded

On Thu, Mar 4, 2021 at 10:28 RADERMACHER Ansgar via lttng-dev <lttng-dev@lists.lttng.org<mailto:lttng-dev@lists.lttng.org>> wrote:
Hi,

when doing tracing with barectf, the trace elements are written into a buffer first and only written when the buffer is full - or if the function barectf_platform_fs_fini gets called.
In case of a crash, it's therefore possible to loose some events. What is the best option to prevent this issue? I've used signal handlers that call the function barectf_platform_fs_fini. This seems to work well with Linux, but is not portable. When using the provided sample platform, there is also the option to reduce the buffer size, but this is not ideal, as trace events that are too big for the buffer are not written.

First question: why are you using barectf on Linux vs LTTng?

LTTng supports writing the sub-buffers to NVRAM and, if the system crashes, convert those sub-buffers to a CTF trace with the lttng-crash utility. See <
https://lttng.org/docs/v2.12/#doc-persistent-memory-file-systems>.

With NVRAM, you could do this with barectf with a corresponding platform using it. It's always up to the platform author to decide where buffers are and where complete packets go; it's not a feature of barectf as such.

Hope it helps,

Phil


Best regards

Ansgar


_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org<mailto:lttng-dev@lists.lttng.org>
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
--
Philippe Proulx

[-- Attachment #1.2: Type: text/html, Size: 4594 bytes --]

[-- Attachment #2: Type: text/plain, Size: 156 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev