Re: [DRE-maint] unicorn: native systemd service

[Christos Trochalakis <yatiohi@ideopolis.gr>]

On Sat, Jun 27, 2015 at 03:05:01AM +0000, Eric Wong wrote:
>Christos Trochalakis <yatiohi@ideopolis.gr> wrote:
>> On Thu, Jun 25, 2015 at 11:26:26PM +0000, Eric Wong wrote:
>> >With socket activation, you should just be able to kill unicorn using
>> >SIGQUIT (just master, or even all workers) and restart without ever
>> >dropping a connection.  I do NOT suggest using SIGTERM for unicorn,
>> >since that'll cause the master to kill all workers ASAP.
>>
>> Yes, you are right socket activation is also an option! I have made some
>> experiments with a simple rack app to test it.
>>
>> systemd uses the LISTEN_FDS env variable that is an integer indicating the
>> number of inherited file descriptors. Those FDs have consecutive numbers
>> starting from `SD_LISTEN_FDS_START` which is `3` (man sd_listen_fds).
>>
>> So for example if LISTEN_FDS="2", UNICORN_FD should be "3,4". I used a
>> simple wrapper script for that. Here is the full configuration:
>
>OK, I'll probably add LISTEN_FDS and LISTEN_PID support to unicorn
>directly so the wrapper is unnecessary.
>
><snip>
>
>> KillMode=mixed
>
>I don't think KillMode=mixed is necessary, here.  systemd can send
>SIGQUIT to workers.
>

Perhaps there is a race here, if a worker receives SIGQUIT first, and the
master respawns a new worker before receiving/handling its own SIGQUIT.
This is definitely a longshot, and will probably never happen, but
even then every process will eventually quit gracefully.

><snip>
>
>> TCP socket options are not applied by unicorn on inherited sockets (TCPSocket
>> === sock is false). systemd socket files have support for most options now but
>> we might want unicorn to `setsockopt` them as well. For example,
>> 'DeferAcceptSec', 'KeepAliveIntervalSec', 'NoDelay' are supported since v216, so
>> they are not available in jessie (v215).
>
>They are now :)
>
>http://bogomips.org/unicorn-public/m/1435373879-4299-1-git-send-email-e@80x24.org.txt
>
>We don't have KeepAliveIntervalSec
>(aka TCP_KEEPINTVL) since it's not-portable and probably over
>Are you sure about that?

I added 'KeepAliveIntervalSec' by mistake. `KeepAlive`, a supported
option by unicorn, is included in jessie (systemd v215). Either way with
your last patch those options are enforced by unicorn.

>
>> socket activation is a really interesting setup, but personally I would not run
>> it with a large application. Clients would have to wait for the new master to
>> be up and running before a reply is returned, and that could take tenths of
>> seconds. The USR2 rexec solves that problem since both old and new workers are
>> accepting on the socket and we can kill the old ones when we are ready. In that
>> case the PIDFile trick is handy to support zero downtime restarts with no
>> latency.
>
>Is it possible to make systemd fire up two unicorn masters?
>That would be a nice feature to have with socket activation.

That would be great! I am a bit surprised that specifying multiple
services ('Service=' option) is not supported.