Willy,

Am 28.05.2018 um 14:59 schrieb Willy Tarreau:
> OK this was fine. I noticed that you also emitted the value for the servers
> but that this one was not filled in the code, and that there was almost
> nothing to do to get it, for the response case, which is useful because if
> you see that one server has many more errors than the other ones, you know
> it emits some different contents (eg: looping cookies).  I just had to add
> the following lines for this :
> 
>           if (objt_server(s->target))
>                    
> HA_ATOMIC_ADD(&objt_server(s->target)->counters.failed_rewrites, 1);

Understood. I just grepped through the code to find references of other
counters and copied some good-looking code over. Maybe all those counter
incrementing locations should be unified into an (inlined?) function
which automatically does the right thing?

> Also I found that the #%u with the request ID was missing for the request
> log, it was only for the response, so I fixed it. It works well and seems
> pretty useful. I'll check how to fit it into the stats page. It'll probably
> be OK in the tooltip which already contains the distribution by status code.
> 

The joys of copy and paste. Thank you. I also noticed that you adapted
the stats page already.

As I did not note anything in the commit message: It should be discussed
whether this patch(es) should be backported. It looks fairly safe too me
and could be helpful for people running 1.8 (or even older).

Best regards
Tim Düsterhus

Reply via email to