[systemd-devel] rpcbind static, want enabled

[Felix Miata]

Andrei Borzenkov composed on 2014-10-21 12:12 (UTC+0400):

> Felix Miata wrote:

>> Andrei Borzenkov composed on 2014-10-21 11:29 (UTC+0400):

>>> Felix Miata wrote:

>>>> I have 27 Fedora 21 & 22 installations to real hardware, all originating via
>>>> HTTP process. Half work as expected. Those that do have NetworkManager not
>>>> installed, and have

>>>>         rpcbind.service                         enabled

>>>> Those that misbehave by having a more than a minute delay on attempt to mount
>>>> nfs via fstab entries, of which half do and half do not have NetworkManager
>>>> installed, have

>>>>         rpcbind.service                         static

>>> then it is socket activated, you need to enable rpcbind.socket.

>> But every one of the failing installations already had:

>>         rpcbind.socket                          enabled

>> same as the working installations.

>>>> With chkconfig and sysvinit it was a simple matter to enable rpcbind. With
>>>> systemd evolved to v216 it seems I should be able to 'systemctl enable
>>>> rpcbind.service', but that changes nothing on these installations. What's the
>>>> obstacle here? If the explanation is in man systemctl or man rpcbind, I'm not
>>>> finding it.

>> Lots more blame, unit status and logs at:
>> http://fm.no-ip.com/Tmp/Linux/F/NFS/

>> Also Fedora mailing list thread:
>> https://lists.fedoraproject.org/pipermail/test/2014-May/121461.html
 
> May 29 00:40:00 g5eas rpc.statd[692]: Version 1.3.0 starting
> May 29 00:40:00 g5eas systemd[1]: Started RPC bind service.
 
> It looks like rpcbind is correctly attempted to be started when statd
> tries to contact it so socket activation appears to work as far as
> systemd is concerned. Then for some reasons rpcbind fails to process
> request from statd.
 
> Enabling more debugging on rpcbind and/or statd may explain why
> registration fails.

Those are not things I know how or wish to pursue. I found a workaround, no thanks
to the systemctl 216-10.fc22 man page, which says:

	"enable NAME...

    Enable one or more unit files or unit file instances, as specified on the command line...."

That's invalid WRT rpcbind. In order to enable rpcbind I found the following
produces satisfactory results:

	systemctl add-wants multi-user.target rpcbind

The question remains how and why 13 of 26 (I miscounted in my original thread post)
installations were set to static instead of enabled in the first place, and whether
the workaround amounts to an optimal solution.
-- 
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/