Thanks for the report.
Can you inspect what consumes the 10 seconds? I think strace can be
used for the purpose.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
Ok, thanks Tatsuo,
I testet the latest code (the one of
Ok, thanks Tatsuo,
I testet the latest code (the one of 2010-04-07), and it is working fine,
also in the case of the unpluged wire, yet if it lasts more than 10
seconds to disconnect(and continue normaly), and it would be desiderable to
be zero. Anyway, the improvement it's been so good.
Thanks
I have committed the change (using child_exit(0)). At first it seems
it's enough to check exit_request flag but consider following
scenario:
(1)-
(2)- fds = select(num_fds, readmask, writemask, exceptmask, NULL);
if (fds == -1)
{
(3)- if (errno == EINTR)
Thanks again Tatsuo,
I've been giving it a try, and, it seems, as you say, that with exit(0)
something goes wrong with the recovery, but, agains as you say, it seems
solved with child_exit(0), maybe it lasts some more seconds to continue
after unplug the wire, but I cannot assure that, anyway, it
Thanks again Tatsuo,
You are welcome!
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
I've been giving it a try, and, it seems, as you say, that with exit(0)
something goes wrong with the recovery, but, agains as you say, it
This will miss some house keeping jobs such as count down connection
counter. Thus you will be in trouble when you do online recovery. You
would want to use child_exit(0) instead exit(0).
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:
Hi Tatsuo,
We have been doing some more debugging, and we found what the problem is. As
I commented in my previous message, the problem was that a proccess (where
the client where connecting), didn't die.
We have found a solution,... it is clearly not an elegant solution, but it
works for us (I
Daniel,
I have committed fix for this. Please check it out.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
Daniel,
Looks like interrupted connect() ought to check if SIGQUIT has been
delivered. Will check. I have no time
Hi Tatsuo,
I did some more debug, and it seems the problem I had is solved. But there
is still a problem.
The test I did is:
1) Start postgres node2
2) Start postgres node1
3) start pgpool (node1)
4) From a remote terminal (client1):
for i in $(seq 0 2000); do { echo $(date +%H:%M:%S); psql -d
Daniel,
Looks like interrupted connect() ought to check if SIGQUIT has been
delivered. Will check. I have no time until next week though.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
Hi Tatsuo,
I did some more debug, and
Thanks to you Tatsuo,...
I am glad my report helped you! As soon as you tell me it is done I will
start testing/debugging it again.
2010/3/2 Tatsuo Ishii is...@sraoss.co.jp
Daniel,
Thanks for the report!
I spoke in another message about this problem, yet, I debugged deeper and
I
have
Daniel,
I have committed fix into CVS HEAD. Could you try it out?
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp
Thanks to you Tatsuo,...
I am glad my report helped you! As soon as you tell me it is done I will
start
Daniel,
Thanks for the report!
I spoke in another message about this problem, yet, I debugged deeper and I
have more specific information, that, maybe, can be usefull.
(The thread I spoke something about was:
http://lists.pgfoundry.org/pipermail/pgpool-general/2010-February/002565.html
)
13 matches
Mail list logo