Nut upsuser - Jul 2015 - upsd not starting sometimes (Porteus 3.1, nut 2.7.2)

I am running tests on my system and UPS, making sure that it is reliably able to
come up, detect power loss, shutdown safely, and then come back up when the
power returns.

It does that MOST of the time.  However, a significant part of the time, the
system comes up, and then doesn't respond to loss of power.  Doing some
checking, I find that the reason is because upsd never started.  Capturing its
output, I see that it says :

Fatal error: A previous upsd instance is already running!
Either stop the previous instance first, or use the 'reload' command.

Again note that this only happens SOME of the time....all other times, all 3
things get started that are supposed to (upsdrvctl, upsd, upsmon).  Even when
upsd fails, the other two services start.

This is being run in Porteus 3.1 with nut 2.7.2.  I'm calling my script from
/etc/rc.d/rc.local.  This script contains:

#!/bin/sh
/usr/local/ups/sbin/upsdrvctl -u ups start
sleep 2
/usr/local/ups/sbin/upsd  -u ups
sleep 2
/usr/local/ups/sbin/upsmon -u ups


I checked that upsd wasn't getting started from anywhere else, and I looked
in the NUT code to see if I could figure out why it thought upsd was already
started, but nothing was obvious.

Does anyone have any other suggestions for tracking down why upsd errors out
like that sometimes?

Sincerely,
Rob Groner
Software Engineer

RTD Embedded Technologies, Inc.
ISO 9001 and AS9100 Certified
Ph: 814-234-8087
www.rtd.com<http://www.rtd.com/>

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.alioth.debian.org/pipermail/nut-upsuser/attachments/20150707/1ee9ea1b/attachment.html>

On Jul 7, 2015, at 4:12 PM, Rob Groner <rgroner at RTD.com> wrote:
> It does that MOST of the time.  However, a significant part of the time,
the system comes up, and then doesn?t respond to loss of power.  Doing some
checking, I find that the reason is because upsd never started.  Capturing its
output, I see that it says :
>  
> Fatal error: A previous upsd instance is already running!
> Either stop the previous instance first, or use the 'reload'
command.
What might be happening is that the PID file is left over from the last test,
and another process got that PID instead.

Due to history and conflicting requirements, the path selection is not as simple
as it could be. From ./configure --help:

  --with-statepath=PATH   path for ups state files (/var/state/ups)
  --with-altpidpath=PATH  path for driver/upsd .pid files (<statepath>)
[...]
  --with-pidpath=PATH     path for .pid files (/var/run)

You might not want to use the default for --with-altpidpath: /var/state/* is not
generally cleared at boot time, whereas /var/run is either wiped or mounted on a
RAM filesystem.

However, you might want the statepath protected - Debian uses the following:

$ ls -ld /var/run/nut
drwxrwx--- 2 root nut 60 Jun 24 17:13 /var/run/nut

-- 
Charles Lepple
clepple at gmail

On Tue, 7 Jul 2015, Rob Groner wrote:
> This is being run in Porteus 3.1 with nut 2.7.2.
My first thought was that this is more systemd wierdness, but I believe 
that Porteus is based on Slackware which doesn't use systemd - is that 
correct?  Roger

Correct, Porteus is based on Debian.  

I'm trying Charles' advice, initially just by making sure the /var/state
dir is empty on boot.  That's pretty easy to do with Porteus since you can
prevent any permanent changes to the file system.

Sincerely,
Rob Groner
Software Engineer

RTD Embedded Technologies, Inc.
ISO 9001 and AS9100 Certified
Ph: 814-234-8087
www.rtd.com

-----Original Message-----
From: Nut-upsuser [mailto:nut-upsuser-bounces+rgroner=rtd.com at
lists.alioth.debian.org] On Behalf Of Roger Price
Sent: Wednesday, July 08, 2015 4:54 AM
To: nut-upsuser Mailing List
Subject: Re: [Nut-upsuser] upsd not starting sometimes (Porteus 3.1, nut 2.7.2)

On Tue, 7 Jul 2015, Rob Groner wrote:
> This is being run in Porteus 3.1 with nut 2.7.2.
My first thought was that this is more systemd wierdness, but I believe that
Porteus is based on Slackware which doesn't use systemd - is that correct? 
Roger

_______________________________________________
Nut-upsuser mailing list
Nut-upsuser at lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/nut-upsuser

That seems to have fixed it...I've run it twice now through our tests and no
recurrence of the problem.

So I'll use the --with-altpidpath configure option to move it to /var/run
instead of /var/state.

Thanks!

Sincerely,
Rob Groner
Software Engineer

RTD Embedded Technologies, Inc.
ISO 9001 and AS9100 Certified
Ph: 814-234-8087
www.rtd.com

-----Original Message-----
From: Charles Lepple [mailto:clepple at gmail.com] 
Sent: Tuesday, July 07, 2015 8:25 PM
To: Rob Groner
Cc: nut-upsuser Mailing List
Subject: Re: [Nut-upsuser] upsd not starting sometimes (Porteus 3.1, nut 2.7.2)

On Jul 7, 2015, at 4:12 PM, Rob Groner <rgroner at RTD.com> wrote:
> It does that MOST of the time.  However, a significant part of the time,
the system comes up, and then doesn't respond to loss of power.  Doing some
checking, I find that the reason is because upsd never started.  Capturing its
output, I see that it says :
>  
> Fatal error: A previous upsd instance is already running!
> Either stop the previous instance first, or use the 'reload'
command.
What might be happening is that the PID file is left over from the last test,
and another process got that PID instead.

Due to history and conflicting requirements, the path selection is not as simple
as it could be. From ./configure --help:

  --with-statepath=PATH   path for ups state files (/var/state/ups)
  --with-altpidpath=PATH  path for driver/upsd .pid files (<statepath>)
[...]
  --with-pidpath=PATH     path for .pid files (/var/run)

You might not want to use the default for --with-altpidpath: /var/state/* is not
generally cleared at boot time, whereas /var/run is either wiped or mounted on a
RAM filesystem.

However, you might want the statepath protected - Debian uses the following:

$ ls -ld /var/run/nut
drwxrwx--- 2 root nut 60 Jun 24 17:13 /var/run/nut

--
Charles Lepple
clepple at gmail