Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: neutral (athena.apache.org: local policy)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1077)
Subject: Re: Stopping externals
From: Damien Katz <damien@apache.org>
In-Reply-To: <ec7a93a1002121305w4f81bceai10937d3f3dde4842@mail.gmail.com>
Date: Fri, 12 Feb 2010 13:09:19 -0800
Content-Transfer-Encoding: quoted-printable
Message-Id: <C8BB4AAF-71BF-4642-A33B-2E561C1C40B8@apache.org>
References: <e2111bbb1002021103v13c0ce19gab3ffd41d640a9ee@mail.gmail.com>
 <e2111bbb1002021103hf198eeoefdd4facd5576da5@mail.gmail.com>
 <201002121326.57128.markus@buyways.nl>
 <ec7a93a1002121305w4f81bceai10937d3f3dde4842@mail.gmail.com>
To: user@couchdb.apache.org

Haven't been following the whole discussion, but if the erlang vm is =
crashed and restarted, then the child process are not forcibly killed. =
So it's up to the child processes to terminate themselves in that case, =
even with the spawnkillable stuff.

-Damien


On Feb 12, 2010, at 1:05 PM, Randall Leeds wrote:

> It looks like lounge is *almost* doing the right thing.
> Couchspawnkillable should be killing it, so it's strange that it hangs =
around.
>=20
> I added a little fix that should try to signal the background thread
> to stop when sys.in shut downs.
> Give the version on this new branch a try, please:
> =
http://github.com/tilgovi/couchdb-lounge/raw/replicatorshutdown/replicator=
/replication_notifier.py
>=20
> Randall
>=20
> On Fri, Feb 12, 2010 at 04:26, Markus Jelsma <markus@buyways.nl> =
wrote:
>> Hi Paul,
>>=20
>>=20
>> CouchDB 0.10 does seem to have couchspawnkillable;
>> couch-1 [~/apache-couchdb-0.10.0] # find  -name "couchspawnkillable*"
>> ./src/couchdb/priv/couchspawnkillable_win.c
>> ./src/couchdb/priv/couchspawnkillable
>> ./src/couchdb/priv/couchspawnkillable.sh
>> ./src/couchdb/priv/.deps/couchspawnkillable_win.Po
>>=20
>> Can you confirm this is the desired output?
>>=20
>> Anyway, if so there might probably be a bug then because the =
replicator does
>> not quit with CouchDB. I am not entirely sure whether the replicator =
checks
>> for a terminating pipe but i guess it does.
>>=20
>> Here is the Python code that sits in the main loop:
>> stuff =3D sys.stdin.readline()
>>    if not stuff:
>>        return
>>=20
>> Perhaps a lounge developer can tell us whether the replicator =
actually behaves
>> as Troy suggested.
>>=20
>>=20
>> Cheers,
>>=20
>>=20
>>> Don't recall when that came in. To check:
>>>=20
>>> $ find $PREFIX/lib/couchdb -name "couchspawnkillable*"
>>>=20
>>> If that finds something then it should be new enough to =
theoretically
>>> kill things.
>>>=20
>>> Also, you should still use Troy's pattern for reading from stdin to =
be
>>> well behaved for the case that we have a bug. :)
>>>=20
>>> HTH,
>>> Paul Davis
>>>=20
>>> Managed to forget the ML in my reply.
>>=20
>>=20
>> Markus Jelsma - Technisch Architect - Buyways BV
>> http://www.linkedin.com/in/markus17
>> 050-8536620 / 06-50258350
>>=20
>>=20