Tarantool development patches archive
 help / color / mirror / Atom feed
From: Alexander Turenko <alexander.turenko@tarantool.org>
To: Maxim Melentiev <dmarc-noreply@freelists.org>
Cc: Konstantin Osipov <kostja@tarantool.org>,
	tarantool-patches@freelists.org
Subject: [tarantool-patches] Re: [PATCH] tarantoolctl doesn't fail when box.cfg is delayed in init script
Date: Mon, 19 Aug 2019 18:05:32 +0300	[thread overview]
Message-ID: <20190819150532.gh25tyc73mcfazeg@tkn_work_nb> (raw)
In-Reply-To: <59464C78-29E6-4F91-B8E9-CA01CB63FE65@corp.mail.ru>

So, in brief the patch can be described in the following way?

 | tarantoolctl patches (wraps) box.cfg function two times: before an
 | init script to set default values from
 | /etc/{default,sysconfig}/tarantool and after the init script to
 | discard changing of a pid file during an instance work time.
 |
 | The second patching procedure fails if an instance file did not call
 | box.cfg{} before an init script ends. Other then that an instance,
 | which calls box.cfg{} in, say, a background fiber, works almost fine
 | (see below).
 |
 | The patch moves the second patching procedure into the box.cfg
 | wrapper created during the first patching. So the second patching
 | procedure is called only after box.cfg{} was invoked second or next
 | time, so it does not fails anymore as it did before.
 |
 | However there are the following relatively minor flaws for
 | applications that invoked box.cfg{} in background:
 |
 | <native tarantoolctl / systemd differences here>

I don't insist on my wording, but want to verify my understanding.

If it is so, then I don't have objections against the patch. You need
however pass the formal process: file an issue, ask one of maintainers
to set it to a right milestone (if it is 1.10 your business need should
be really important), fix the issue from the commit message, help
documentation team with documenting the new feature right when the patch
will be merged (they maybe will ask you something, maybe don't -- but I
suggest to monitor the tarantool/doc issue when it will be created).

BTW, is this is enough for you in the scope of your task or we need to
boost other tarantoolctl discussions very soon?

See other comments below.

WBR, Alexander Turenko.

On Mon, Aug 19, 2019 at 12:48:22PM +0300, Maxim Melentiev wrote:
> From cd94e49e7b23a30ed05e574eb8441b477fc1c47a Mon Sep 17 00:00:00 2001
> From: Max Melentiev <m.melentiev@corp.mail.ru>
> Date: Mon, 19 Aug 2019 10:35:55 +0300
> Subject: [PATCH] tarantoolctl: allow to start instances without a box.cfg{}

It is a bit confusing: AFAIU it allows to call box.cfg{} in background
after an init script execution. But it does not allow to miss box.cfg{}
at all.

> 
> Before this patch, tarantoolctl would fail if box.cfg{} was not
> called in an instance script. It used too patch box.cfg
> once again after init script to prevent chainging pid file.

Typo: chainging.

> 
> This patch allows to start an instance without immediate call to
> box.cfg{} in it. For example, managed instances which receive
> configuration from external server.
> 
> @TarantoolBot document
> Title: tarantoolctl allows to start instances without a box.cfg{}
> 
> tarantoolctl now works for instances without box.cfg{} or
> with dealyed box.cfg{} call. This can be managed instances which receive

Typo: dealyed.

> configuration from external server.
> 
> For such instances `tarantoolctl start` goes to background when
> box.cfg{} is called, so it will wait until options for box.cfg are received.
> However this is not the case for daemon management systems like systemd,
> as they handle bockgrounding on their side.

Typo: bockgrounding.

> ---
>  extra/dist/tarantoolctl.in | 21 +++++++++------------
>  1 file changed, 9 insertions(+), 12 deletions(-)
> 
> diff --git a/extra/dist/tarantoolctl.in b/extra/dist/tarantoolctl.in
> index 8adb57533..fd1abf9dc 100755
> --- a/extra/dist/tarantoolctl.in
> +++ b/extra/dist/tarantoolctl.in
> @@ -483,6 +483,15 @@ local function wrapper_cfg(cfg)
>          os.exit(1)
>      end
> 
> +    local box_cfg_mt = getmetatable(box.cfg)
> +    local orig_cfg_call = box_cfg_mt.__call
> +    box_cfg_mt.__call = function(old_cfg, new_cfg)
> +        if old_cfg.pid_file ~= nil and new_cfg ~= nil and new_cfg.pid_file ~= nil then
> +            new_cfg.pid_file = old_cfg.pid_file
> +        end
> +        orig_cfg_call(old_cfg, new_cfg)
> +    end
> +

Once you are here it would be good to add a comment what this block
does.


>      return data
>  end
> 
> @@ -547,18 +556,6 @@ local function start()
>          end
>          os.exit(1)
>      end
> -    local box_cfg_mt = getmetatable(box.cfg)
> -    if box_cfg_mt == nil then
> -        log.error('box.cfg() is not called in an instance file')
> -        os.exit(1)
> -    end
> -    local old_call = box_cfg_mt.__call
> -    box_cfg_mt.__call = function(old_cfg, cfg)
> -        if old_cfg.pid_file ~= nil and cfg ~= nil and cfg.pid_file ~= nil then
> -            cfg.pid_file = old_cfg.pid_file
> -        end
> -        old_call(old_cfg, cfg)
> -    end
>      return 0
>  end
> 
> --
> 2.21.0
> 
> 
> > On 19 Aug 2019, at 10:54, Konstantin Osipov <kostja@tarantool.org> wrote:
> > 
> > * Max Melentiev <dmarc-noreply@freelists.org> [19/08/19 10:40]:
> >> Before this patch tarantoolctl failed with error if box.cfg
> >> was not called in init script because it used too patch box.cfg
> >> once again after init script. I've changed it to patch box.cfg
> >> second time inside wrapper_cfg.
> > 
> > Please explain better what this patch does, not how you achieved
> > that.
> > 
> > And while you are at it, explain why we need a secondary patching
> > of box.cfg at all. What's the point of preserving the pid file?
> > 
> > tarantoolctl: allow to start instances without a box.cfg{}
> > 
> > Before this patch, tarantoolctl would fail if box.cfg{} was not
> > called in an instance script.
> > 
> > This patch allows to start an instance without box.cfg{} in it. 
> > Such an instance:
> > - may only be started under systemd? 
> > - what can it do? 
> > - ???
> > 
> > Please add a docboc entry, since it 's a significant behaviour
> > change. How such an instance is used? 
> > 
> > Otherwise LTGM.
> > 
> > 
> >> ---
> >> extra/dist/tarantoolctl.in | 21 +++++++++------------
> >> 1 file changed, 9 insertions(+), 12 deletions(-)
> >> 
> >> diff --git a/extra/dist/tarantoolctl.in b/extra/dist/tarantoolctl.in
> >> index 8adb57533..fd1abf9dc 100755
> >> --- a/extra/dist/tarantoolctl.in
> >> +++ b/extra/dist/tarantoolctl.in
> >> @@ -483,6 +483,15 @@ local function wrapper_cfg(cfg)
> >>         os.exit(1)
> >>     end
> >> 
> >> +    local box_cfg_mt = getmetatable(box.cfg)
> >> +    local orig_cfg_call = box_cfg_mt.__call
> >> +    box_cfg_mt.__call = function(old_cfg, new_cfg)
> >> +        if old_cfg.pid_file ~= nil and new_cfg ~= nil and new_cfg.pid_file ~= nil then
> >> +            new_cfg.pid_file = old_cfg.pid_file
> >> +        end
> >> +        orig_cfg_call(old_cfg, new_cfg)
> >> +    end
> >> +
> >>     return data
> >> end
> >> 
> >> @@ -547,18 +556,6 @@ local function start()
> >>         end
> >>         os.exit(1)
> >>     end
> >> -    local box_cfg_mt = getmetatable(box.cfg)
> >> -    if box_cfg_mt == nil then
> >> -        log.error('box.cfg() is not called in an instance file')
> >> -        os.exit(1)
> >> -    end
> >> -    local old_call = box_cfg_mt.__call
> >> -    box_cfg_mt.__call = function(old_cfg, cfg)
> >> -        if old_cfg.pid_file ~= nil and cfg ~= nil and cfg.pid_file ~= nil then
> >> -            cfg.pid_file = old_cfg.pid_file
> >> -        end
> >> -        old_call(old_cfg, cfg)
> >> -    end
> >>     return 0
> >> end
> > 
> > -- 
> > Konstantin Osipov, Moscow, Russia
> 
> 

  reply	other threads:[~2019-08-19 15:05 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-19  7:39 [tarantool-patches] " Max Melentiev
2019-08-19  7:54 ` [tarantool-patches] " Konstantin Osipov
2019-08-19  9:48   ` Maxim Melentiev
2019-08-19 15:05     ` Alexander Turenko [this message]
2019-08-20 11:24       ` Maxim Melentiev
2019-08-20 11:55         ` Konstantin Osipov
2019-08-20 23:15         ` Alexander Turenko
2019-08-22 13:06           ` Kirill Yukhin
2019-08-20 11:54     ` Konstantin Osipov
2019-08-20 12:08   ` Maxim Melentiev

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=20190819150532.gh25tyc73mcfazeg@tkn_work_nb \
    --to=alexander.turenko@tarantool.org \
    --cc=dmarc-noreply@freelists.org \
    --cc=kostja@tarantool.org \
    --cc=tarantool-patches@freelists.org \
    --subject='[tarantool-patches] Re: [PATCH] tarantoolctl doesn'\''t fail when box.cfg is delayed in init script' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox