Hi, ZhiDao.
If master dies often as you said, I believe
community shell treat it as a bug and fix it ASAP.
Personally I won't meet it frequently. Only
at when system OOM or user misoperation. Thanks.
___
nginx-devel mailing list
nginx-devel@nginx.org
Hello!
On Sat, Oct 08, 2016 at 06:31:02PM +1100, Rob N ★ wrote:
> # HG changeset patch
> # User Rob N ★
> # Date 1475910300 -39600
> # Sat Oct 08 18:05:00 2016 +1100
> # Node ID 205f2148260460379f9b0889cdb8015994028c73
> # Parent 1606a817c1d48ed351f1dd7d9cb9c996e2c77b9a
Hi, Maxim. When master OOM, system daemon program like
supervisord will trying to pull it up. but it can’t be
done because of worker binds the port. And i believe the master process shell
handler all the resources. Slaves shell not exist on its own.
Hello!
It's a common case that all the worker processes are still live while the
master process isn't exist.
So I think nginx should continue to work as Maxim said.
Btw, we noticed that nginx.pid is deleted when we do the upgrade command.
kill -USR2 nginx.pid; sleep(2); kill -HUP nginx.oldpid;
Hello!
On Thu, Oct 13, 2016 at 08:38:56AM +0800, 李林哲 wrote:
> This patch is solving the problem that the sub process (worker
> process) will not quit when master process quits on SIGKILL.
> which will be done by OS (especially Linux) when system meet
> OOM.
Why do you think that worker
# HG changeset patch
# User lilinzhe
# Date 1476321680 -28800
# Node ID d13daa14b1df9e04d7bf339c1669740edf19de0e
# Parent 7cdf69d012f02a80c94d930b29c71847e203e4d6
worker process: make sure to quit when master quits
when theres prctl(PR_SET_PDEATHSIG). make worker to use it.
on
Thanks for reply.
This patch is solving the problem that the sub process (worker process) will
not quit when master process quits on SIGKILL. which will be done by OS
(especially Linux) when system meet OOM.
I've found a bug, and will send a new patch for comment.
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
On 2016-10-12 16:33, Maxim Dounin wrote:
> [...]
>
> From package maintaining point of view it should be a good idea to
> add the option to nginx builds. This will allow others to easily
> compile dynamic modules to be loaded into nginx shipped
details: http://hg.nginx.org/njs/rev/8046b999aaaf
branches:
changeset: 198:8046b999aaaf
user: Igor Sysoev
date: Wed Oct 12 19:08:57 2016 +0300
description:
A fix in Array.slice() function.
diffstat:
njs/njs_array.c | 34 +-
Dear list,
I am interested in using this data structure.
Would it be possible to have a few usage examples of it?
I looked in the nginx code via github but found only
few usages and not very informative about how to use it properly.
I am interested in those functions:
ngx_radix_tree_t
Hello!
On Wed, Oct 12, 2016 at 03:58:25PM +0200, Thomas Deutschmann wrote:
> nginx-1.11.5 introduced a new configure option "--with-compat" however
> I don't find any documentation regarding this option.
>
> Could you please give us some insight when you expect that this option
> should be used
Hello!
On Wed, Oct 12, 2016 at 08:44:25PM +0800, 李林哲 wrote:
> # HG changeset patch
> # User lilinzhe
> # Date 1476275566 -28800
> # Node ID 07d294d39cc6ec07769c1a72d5ee9b7da27b9ce1
> # Parent 7cdf69d012f02a80c94d930b29c71847e203e4d6
> fork: subprocess quit when parent died
>
>
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Hi,
nginx-1.11.5 introduced a new configure option "--with-compat" however
I don't find any documentation regarding this option.
Could you please give us some insight when you expect that this option
should be used (especially from distribution
2016-10-12 14:09 GMT+02:00 Roman Arutyunyan :
> Do you have this issue without slice?
No, if I send a the same request that slice module would send to
upstream through nginx
without slice activated, I do not see the same errors
> Anyway, it would be nice to see debug log
Working
On Wednesday 12 October 2016 15:05:48 George . wrote:
> Hi Valentin,
>
> Than you for your fast and correct reply.
>
> Yes, you are right, it was oversight that r->out->buf if taken from
> pipe->pool, which is actually r->pool.
> What if we pfree this bufs from r->pool? Who holds the reference
# HG changeset patch
# User lilinzhe
# Date 1476275566 -28800
# Node ID 07d294d39cc6ec07769c1a72d5ee9b7da27b9ce1
# Parent 7cdf69d012f02a80c94d930b29c71847e203e4d6
fork: subprocess quit when parent died
diff -r 7cdf69d012f0 -r 07d294d39cc6 auto/os/linux
--- a/auto/os/linux Tue Oct
Hi,
On Tue, Oct 11, 2016 at 08:26:29PM +0200, Bjørnar Ness wrote:
> Hello nginx-devel
>
> I am seeing lots of 'upstream sent more data than specified in
> "Content-Length"' errors
> when using nginx with proxy_cache and slice module. The relevant
> nginx.conf settings are:
>
>
> > slice 1m;
>
Hi Valentin,
Than you for your fast and correct reply.
Yes, you are right, it was oversight that r->out->buf if taken from
pipe->pool, which is actually r->pool.
What if we pfree this bufs from r->pool? Who holds the reference after
sending these buffers (except may be in NGX_THREADS version)?
Hi,
Is your origin server something like Apache + mod_fastcgi + mod_deflate ?
I did not see this specifically with Nginx but you may hit the same issue
described here: https://ma.ttias.be/apaches-mod_fastcgi-mod_deflate-troubles/
Best regards,
Thierry
-Message d'origine-
De :
On Wednesday 12 October 2016 13:22:08 George . wrote:
> Hi all,
>
> What do you think about freeing buffers after being sent by send_chain?
> Because we have a special case were we use 4 megabyte buffer instead of
> nginx default 8 * 8 K and we found that these buffers are actually freed
> after
Hi all,
What do you think about freeing buffers after being sent by send_chain?
Because we have a special case were we use 4 megabyte buffer instead of
nginx default 8 * 8 K and we found that these buffers are actually freed
after request had been finished, so if there hundreds of slow downstream
21 matches
Mail list logo