Xen devel - Jul 2011 - [PATCH] Fix xend restarting problem on NetBSD

Hello,

This patch fixes a problem with xend restart on NetBSD (the rc.d
script was unable to find the pid of the process, so it was not
restarted).

Regards, Roger.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Roger Pau Monné writes ("[Xen-devel] [PATCH] Fix xend restarting problem on
NetBSD"):
> This patch fixes a problem with xend restart on NetBSD
Hi.  Thanks for the patch.  But I think this patch does more than you
say it does ?

It introduces a new xen_startcmd, xen_stop and xen_status.  Presumably
these were previously provided by standard NetBSD machinery ?  Why is
that standard machinery unsuitable for start and status ?
> (the rc.d script was unable to find the pid of the process, so it
> was not restarted).
Why wasn''t the standard NetBSD machinery able to find the pid ?
How does its approach differ from check_pidfile ?

Sorry if these are all stupid questions.  If you can refer me to some
documentation about the NetBSD init system I may be able to pretend to
be less ignorant :-).

Thanks,
Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

While searching for a man page to show you the check_pidfile options
I''ve found that you can also set a variable with the path to the
interpreter, so no need to expand the script. The only strange thing
is that I have to use SIGKILL to kill xend, SIGTERM seems to do
nothing (also xend spawns two processes, but I guess this is normal).

Attached an updated patch, with a simpler script.

Thanks for making me look at it again, Roger.

2011/7/15 Ian Jackson <Ian.Jackson@eu.citrix.com>:
> Roger Pau Monné writes ("[Xen-devel] [PATCH] Fix xend restarting
problem on NetBSD"):
>> This patch fixes a problem with xend restart on NetBSD
>
> Hi.  Thanks for the patch.  But I think this patch does more than you
> say it does ?
>
> It introduces a new xen_startcmd, xen_stop and xen_status.  Presumably
> these were previously provided by standard NetBSD machinery ?  Why is
> that standard machinery unsuitable for start and status ?
>
>> (the rc.d script was unable to find the pid of the process, so it
>> was not restarted).
>
> Why wasn''t the standard NetBSD machinery able to find the pid ?
> How does its approach differ from check_pidfile ?
>
> Sorry if these are all stupid questions.  If you can refer me to some
> documentation about the NetBSD init system I may be able to pretend to
> be less ignorant :-).
>
> Thanks,
> Ian.
>

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Roger Pau Monné writes ("Re: [Xen-devel] [PATCH] Fix xend restarting
problem on NetBSD"):
> While searching for a man page to show you the check_pidfile options
> I''ve found that you can also set a variable with the path to the
> interpreter, so no need to expand the script.
Thanks.  This version is much nicer :-).  I still have a couple of
questions.  Please forgive my ignorance about NetBSD.
> +PATH="${PATH}:${SBINDIR}"
> +export PATH
Is SBINDIR not on the path for root''s daemon processes ?  Perhaps we
are doing something else wrong for NetBSD (perhaps we should be using
libexec) ?
> +command_interpreter=`head -n 1 ${command} | awk ''{ print
substr($0,3) }''`
Why does it need to know that ?  I''m happy to believe that it does,
but if so I would have expected to find that there was a ready-made
utility on NetBSD that avoids having to open-code a rune like the one
above.
> The only strange thing is that I have to use SIGKILL to kill xend,
> SIGTERM seems to do nothing
That''s probably a bug in the xend python code.  But I don''t
expect
we''ll try to fix that at this point.
> (also xend spawns two processes, but I guess this is normal).
Yes.
> Attached an updated patch, with a simpler script.
Thanks.  I think my comments so far are small enough (and xend
sufficiently deprecated) that I''ve applied it.  If you would like to
further improve it in the light of my questions above please submit
patches which will apply on top of that one.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Hello,

I''m no NetBSD expert either, it''s the first time I''ve
worked with init
scripts (and NetBSD/xen to tell the truth), so all I''m going to say
it''s what I saw on other scripts or read on the Internet.

2011/7/18 Ian Jackson <Ian.Jackson@eu.citrix.com>:
> Roger Pau Monné writes ("Re: [Xen-devel] [PATCH] Fix xend restarting
problem on NetBSD"):
>> While searching for a man page to show you the check_pidfile options
>> I''ve found that you can also set a variable with the path to
the
>> interpreter, so no need to expand the script.
>
> Thanks.  This version is much nicer :-).  I still have a couple of
> questions.  Please forgive my ignorance about NetBSD.
>
>> +PATH="${PATH}:${SBINDIR}"
>> +export PATH
This was done in start_cmd, so I through it would be better to do it
this way. I think it''s necessary, because it was on the script
already, just moved the place where it was, to make the script
cleaner.
>
> Is SBINDIR not on the path for root''s daemon processes ?  Perhaps
we
> are doing something else wrong for NetBSD (perhaps we should be using
> libexec) ?
>
>> +command_interpreter=`head -n 1 ${command} | awk ''{ print
substr($0,3) }''`
>
> Why does it need to know that ?  I''m happy to believe that it
does,
> but if so I would have expected to find that there was a ready-made
> utility on NetBSD that avoids having to open-code a rune like the one
> above.
Init scripts *should* detect that the daemon is a script and find it''s
parser, but I don''t know why NetBSD fails to detect it
(haven''t looked
at it to know where the problem is). Maybe there''s a better or cleaner
way to do it, but I think that since xend is going to be deprecated
the fix should do for now.
>
>> The only strange thing is that I have to use SIGKILL to kill xend,
>> SIGTERM seems to do nothing
>
> That''s probably a bug in the xend python code.  But I
don''t expect
> we''ll try to fix that at this point.
No problem, SIGKILL works fine.
>
>> (also xend spawns two processes, but I guess this is normal).
>
> Yes.
>
>> Attached an updated patch, with a simpler script.
>
> Thanks.  I think my comments so far are small enough (and xend
> sufficiently deprecated) that I''ve applied it.  If you would like
to
> further improve it in the light of my questions above please submit
> patches which will apply on top of that one.
I''m sorry that my comments my not be very informative, as I said
before I''m new to this stuff. I hope my reasons are good enough.

Thanks, Roger.
>
> Ian.
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Roger Pau Monné writes ("Re: [Xen-devel] [PATCH] Fix xend restarting
problem on NetBSD"):
> I''m no NetBSD expert either, it''s the first time
I''ve worked with init
> scripts (and NetBSD/xen to tell the truth), so all I''m going to
say
> it''s what I saw on other scripts or read on the Internet.
OK, that''ll do :-).  Thanks for your contribution.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel