Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Mart van de Wege]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA256

On Wed, 25 May 2016 20:24:41 +0200
Michael Biebl  wrote:

> Maybe the rpcbind.socket is simply not up yet.
> Elimar et al, could you try the change from
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=812371#31
> 
> Patrik suggested something similar at
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805167#71
> 
> Maybe fixing the dependencies of rpcbind.socket so it can start earlier
> is already sufficient to also fix Elimar's original problem.
> 
> 
> Regards,
> Michael
> 

So far I haven't seen any problems anymore, so for my side I will
consider that change a fix for my problem.

Regards,

Mart
 
-BEGIN PGP SIGNATURE-
Version: GnuPG v2
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=kDX0
-END PGP SIGNATURE-

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Elimar Riesebieter]

* Michael Biebl  [2016-05-25 20:24 +0200]:

> Hi
> 
> On Sun, 15 Nov 2015 13:56:50 +0100 Elimar Riesebieter
>  wrote:
> > Package: rpcbind
> > Version: 0.2.3-0.2
> > Severity: important
> > 
> > Can't run nis:
> > # ypbind -no-dbus -broadcast -debug
> > 1679: add_server() domain: mynisdomain, broadcast
> > 1679: [Welcome to ypbind-mt, version 1.20.1]
> > 1679: ping interval is 20 seconds
> > Cannot register service: RPC: Unable to receive; errno = Connection refused
> > 1679: Unable to register (YPBINDPROG, YPBINDVERS, udp).
> > 
> > Restarting rpc bind gets me back to restart nis successfull.
> > 
> > # service rpcbind restart
> > # service nis restart
> > 
> > Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
> > that this happens only on systemd machines. Running sysvinit doesn't
> > show this behaviour.
> 
> Maybe the rpcbind.socket is simply not up yet.
> Elimar et al, could you try the change from
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=812371#31

This fixed my problem. ypbind is up running with rpcbind 0.2.3-0.2.

Elimar
-- 
  The path to source is always uphill!
-unknown-

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Anders Hammarquist]

In a message of Wed, 25 May 2016 20:24:41 +0200, Michael Biebl writes:
>> Restarting rpc bind gets me back to restart nis successfull.
>> 
>> # service rpcbind restart
>> # service nis restart
>> 
>> Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
>> that this happens only on systemd machines. Running sysvinit doesn't
>> show this behaviour.
>
>Maybe the rpcbind.socket is simply not up yet.

>From what I found, this was not the case. The problem, as I wrote in
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805167#54
is that the distributed rpcbind.socket only listens on unix-domain
sockets, and nis talks udp, and thus won't trigger starting rpcbind 
proper.

/Anders

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Michael Biebl]

Hi

On Sun, 15 Nov 2015 13:56:50 +0100 Elimar Riesebieter
 wrote:
> Package: rpcbind
> Version: 0.2.3-0.2
> Severity: important
> 
> Can't run nis:
> # ypbind -no-dbus -broadcast -debug
> 1679: add_server() domain: mynisdomain, broadcast
> 1679: [Welcome to ypbind-mt, version 1.20.1]
> 1679: ping interval is 20 seconds
> Cannot register service: RPC: Unable to receive; errno = Connection refused
> 1679: Unable to register (YPBINDPROG, YPBINDVERS, udp).
> 
> Restarting rpc bind gets me back to restart nis successfull.
> 
> # service rpcbind restart
> # service nis restart
> 
> Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
> that this happens only on systemd machines. Running sysvinit doesn't
> show this behaviour.

Maybe the rpcbind.socket is simply not up yet.
Elimar et al, could you try the change from
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=812371#31

Patrik suggested something similar at
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805167#71

Maybe fixing the dependencies of rpcbind.socket so it can start earlier
is already sufficient to also fix Elimar's original problem.


Regards,
Michael

-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?



signature.asc
Description: OpenPGP digital signature

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Mark Brown]

On Sun, Nov 29, 2015 at 10:45:39PM +0100, Laurent Bigonville wrote:

> The fact that rpcbind is not automatically started at boot is expected since
> it's now socket activated under systemd.

> It seems that ypbind is using an udp socket instead of the unix socket to
> connect to it

> I'm adding nis/ypbind maintainer  in the loop.

If it's socket activated shouldn't both the UDP and Unix sockets do the
job?


signature.asc
Description: PGP signature

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Laurent Bigonville]

Le 30/11/15 12:09, Mark Brown a écrit :

On Sun, Nov 29, 2015 at 10:45:39PM +0100, Laurent Bigonville wrote:


The fact that rpcbind is not automatically started at boot is expected since
it's now socket activated under systemd.
It seems that ypbind is using an udp socket instead of the unix socket to
connect to it
I'm adding nis/ypbind maintainer  in the loop.

If it's socket activated shouldn't both the UDP and Unix sockets do the
job?

Maybe, not sure though.

Looking at ypbind Fedora package[0], they have added a systemd .service 
file that Requires: rpcbind.service. This should start the rpcbind 
daemon when needed.
IMVHO, adding a systemd .service for all the nis daemon would be great 
and should fix this issue.


If you need help to add systemd support, do not hesitate to ask the 
debian systemd maintainer, they are usually pretty responsive.


Cheers,

Laurent Bigonville

[0] http://pkgs.fedoraproject.org/cgit/ypbind.git/tree/

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Mark Brown]

On Mon, Nov 30, 2015 at 12:37:56PM +0100, Laurent Bigonville wrote:

> Looking at ypbind Fedora package[0], they have added a systemd .service file
> that Requires: rpcbind.service. This should start the rpcbind daemon when
> needed.
> IMVHO, adding a systemd .service for all the nis daemon would be great and
> should fix this issue.

I'm not doing anything with the package until ftpmaster get split
binaries through new and even then I'm not sure this is a good
workaround, I'd expect socket activation to work all the time.


signature.asc
Description: PGP signature

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Laurent Bigonville]

Le 30/11/15 13:07, Mark Brown a écrit :

On Mon, Nov 30, 2015 at 12:37:56PM +0100, Laurent Bigonville wrote:


Looking at ypbind Fedora package[0], they have added a systemd .service file
that Requires: rpcbind.service. This should start the rpcbind daemon when
needed.
IMVHO, adding a systemd .service for all the nis daemon would be great and
should fix this issue.

I'm not doing anything with the package until ftpmaster get split
binaries through new and even then I'm not sure this is a good
workaround, I'd expect socket activation to work all the time.
OK, but i think that in any cases, adding a systemd .service file might 
be a good idea.

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Laurent Bigonville]

On Sun, 15 Nov 2015 13:56:50 +0100 Elimar Riesebieter 
 wrote:


Hi,

> Can't run nis:
> # ypbind -no-dbus -broadcast -debug
> 1679: add_server() domain: mynisdomain, broadcast
> 1679: [Welcome to ypbind-mt, version 1.20.1]
> 1679: ping interval is 20 seconds
> Cannot register service: RPC: Unable to receive; errno = Connection 
refused

> 1679: Unable to register (YPBINDPROG, YPBINDVERS, udp).
>
> Restarting rpc bind gets me back to restart nis successfull.
>
> # service rpcbind restart
> # service nis restart
>
> Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
> that this happens only on systemd machines. Running sysvinit doesn't
> show this behaviour.

The fact that rpcbind is not automatically started at boot is expected 
since it's now socket activated under systemd.


It seems that ypbind is using an udp socket instead of the unix socket 
to connect to it


I'm adding nis/ypbind maintainer  in the loop.

Cheers,

Laurent Bigonville

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Elimar Riesebieter]

Control: retitle -1 "Regression: rpcbind doesn't start at boottime on systemd 
controlled machines."

Does Control accept a '\' at a line break?

Hm
-- 
 We all know Linux is great... it does infinite loops in 5 seconds.
-Linus Torvalds

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Elimar Riesebieter]

Control: retitle -1 "Regression: rpcbind doesn't start at boottime
on systemd controlled machines."

* Elimar Riesebieter  [2015-11-15 13:56 +0100]:

> Package: rpcbind
> Version: 0.2.3-0.2
> Severity: important
> 
> Can't run nis:
> # ypbind -no-dbus -broadcast -debug
> 1679: add_server() domain: mynisdomain, broadcast
> 1679: [Welcome to ypbind-mt, version 1.20.1]
> 1679: ping interval is 20 seconds
> Cannot register service: RPC: Unable to receive; errno = Connection refused
> 1679: Unable to register (YPBINDPROG, YPBINDVERS, udp).
> 
> Restarting rpc bind gets me back to restart nis successfull.

It is not a restart. rpcbind doesn't start at bottime at all!

> # service rpcbind restart
> # service nis restart

Running 

# systemctl start rpcbind
# service nis start

Works as well. nis package doesn't provide a service file.


> Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
> that this happens only on systemd machines.

The systemd systems I've found this behavior are booting via
GRUB option:

GRUB_CMDLINE_LINUX_DEFAULT="init=/lib/systemd/systemd systemd.show_status=true"

> Running sysvinit only doesn't show this behaviour.

Elimar

-- 
 We all know Linux is great... it does infinite loops in 5 seconds.
-Linus Torvalds

Bug#805167: rpcbind: Regression: must restart rpc bind to run ypbind (nis)

[Elimar Riesebieter]

Package: rpcbind
Version: 0.2.3-0.2
Severity: important

Can't run nis:
# ypbind -no-dbus -broadcast -debug
1679: add_server() domain: mynisdomain, broadcast
1679: [Welcome to ypbind-mt, version 1.20.1]
1679: ping interval is 20 seconds
Cannot register service: RPC: Unable to receive; errno = Connection refused
1679: Unable to register (YPBINDPROG, YPBINDVERS, udp).

Restarting rpc bind gets me back to restart nis successfull.

# service rpcbind restart
# service nis restart

Reverting to rpcbind 0.2.1-6.1 solves the problem. Please notice
that this happens only on systemd machines. Running sysvinit doesn't
show this behaviour.

Elimar

-- System Information:
Debian Release: stretch/sid
  APT prefers unstable
  APT policy: (990, 'unstable'), (100, 'experimental')
Architecture: i386 (i686)

Kernel: Linux 4.3.0-gandalf-lxtec-i686 (PREEMPT)
Locale: LANG=de_DE.UTF-8, LC_CTYPE=de_DE.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: sysvinit (via /sbin/init)

Versions of packages rpcbind depends on:
ii  init-system-helpers  1.24
ii  initscripts  2.88dsf-59.2
ii  insserv  1.14.0-5
ii  libc62.19-22
ii  libsystemd0  227-3
ii  libtirpc10.2.5-1
ii  libwrap0 7.6.q-25
ii  lsb-base 9.20150917

rpcbind recommends no packages.

rpcbind suggests no packages.

-- no debconf information