Hello Willy
Sorry for my delay. I was on holidays.
Here truss output.
Scenarios:
1.- start haproxy
$ truss ./haproxy -f prueba.cfg -D -p haproxy13-1.pid >prueba12-1.log 2>&1
2.- start haproxy w/ -sf $(cat...)
$ truss ./haproxy -f prueba.cfg -D -p haproxy13-1.pid >prueba12-2.log
2>&1 -sf $(cat haproxy13-1.pid)
3.- start haproxy w/ -sf $(cat...)
$ truss ./haproxy -f prueba.cfg -D -p haproxy13-1.pid >prueba12-3.log
2>&1 -sf $(cat haproxy13-1.pid)
4.- kill -TTOU. Several times
$ kill -TTOU $(cat haproxy13-1.pid)
5.- kill $(cat...)
kill -TERM $(cat haproxy13-1.pid)
Hope this help,
Later,
Manuel Soto
2008/9/2 Willy Tarreau - w#1wt.eu
<+haproxy+micorreo+479759a914.w#1wt.eu#spamgourmet.com>:
> Hello Manuel,
>
> On Wed, Aug 27, 2008 at 09:10:19AM -0430, haproxy.20.micorreo#spamgourmet.com wrote:
>> Hello List.
>>
>>
>> I've tested in several combinations so I believe that this is a
>> 1.3.15.2 BUG. The same startup/restart script and config file used in
>> 1.2 and 1.3.15.2 but 1.3.x does not works, 1.2 works perfectly.
>
> my problem is that I could not see this problem yet on solaris 8/sparc.
> Could you please run the new process and the old one through truss when
> you do this (provided it's 100% reproducible) ? We need to see what happens
> from the moment the new process fails to bind to ports and the moment it
> reports the problem (with the signal exchanges between both).
>
> I'm sure there's something subtle such as a timing problem for an attempt to
> rebind or something like this, but I fail to see what.
>
> Regards,
> Willy
>
>
>
This archive was generated by hypermail 2.2.0 : 2008/09/30 21:03 CEST