DPDK patches and discussions
 help / color / mirror / Atom feed
From: Stephen Hemminger <stephen@networkplumber.org>
To: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Cc: dev@dpdk.org
Subject: Re: [RFC] Dynamic log/trace control via telemetry
Date: Tue, 16 Aug 2022 19:08:37 -0700	[thread overview]
Message-ID: <20220816190837.40c557bb@hermes.local> (raw)
In-Reply-To: <20220816021738.5498f802@sovereign>

On Tue, 16 Aug 2022 02:17:38 +0300
Dmitry Kozlyuk <dmitry.kozliuk@gmail.com> wrote:

> When debugging a live app, useful info can be obtained from logs or traces
> that were not enabled when it was started and it is undesirable to restart.
> Furthermore, unless the app authors have considered tracing,
> rte_trace_save() is only called on exit, i.e. a shutdown is required again.
> 
> What if the telemetry socket gave the missing control?
> For example:
> 
> --> /eal/log/set_level,debug  
> {"/eal/log/set_level": {"status": "success"}}
> --> /eal/log/set_level,pmd.net.*:debug  
> {"/eal/log/set_level": {"status": "success"}}
> --> /eal/log/set_level,lib.[a-d],debug  
> {"/eal/log/set_level": {"status": "success"}}
> --> /eal/log/set_level,foobar  
> {"/eal/log/set_level": {"status": "fail", "message": "Invalid log level: foobar"}}
> 
> Tracing is more complicated because it requires resource allocation.
> It could be:
> 
> /eal/trace/set_mode,discard
> /eal/trace/set_mode,overwrite
>     rte_trace_mode_set()
> 
> /eal/trace/enable_pattern,<pattern>
> /eal/trace/enable_regex,<regex>
> /eal/trace/disable_pattern,<pattern>
> /eal/trace/disable_regex,<regex>
>     rte_trace_enable_pattern()
>     rte_trace_enable_regex()
> 
> /eal/trace/rearm
>     Clear the trace buffer.
>     Apply the new settings accumulated by the previous commands.
> 
> /eal/trace/save
>     rte_trace_save()
> 
> An open question is how to deal with multi-process.
> Only the primary process listens to the telemetry socket.
> Log and trace settings are not shared between processes.
> OTOH, when enabling some log source, it is desirable to do in all processes.
> 
> In general, telemetry is not for changing the state of the app,
> but logs and traces are diagnostic information that seems to be in-scope.
> A similar suggestion was not opposed recently:
> 
>     http://inbox.dpdk.org/dev/24c49429394294cfbf0d9c506b205029bac77c8b.1657890378.git.anatoly.burakov@intel.com/

Not sure if turning telemetry into a do all control api makes sense.
This seems like a different API.
Also, the default would have to be disabled for application safety reasons.


  parent reply	other threads:[~2022-08-17  2:08 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-15 23:17 Dmitry Kozlyuk
2022-08-17  1:25 ` fengchengwen
2022-08-17  2:08 ` Stephen Hemminger [this message]
2022-08-17 15:15   ` Dmitry Kozlyuk
2022-08-17 15:34     ` Stephen Hemminger
2022-08-17 15:34     ` Morten Brørup
2022-08-20 15:19       ` Dmitry Kozlyuk
2022-08-22  6:47         ` Morten Brørup

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220816190837.40c557bb@hermes.local \
    --to=stephen@networkplumber.org \
    --cc=dev@dpdk.org \
    --cc=dmitry.kozliuk@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).