freeipmi-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Freeipmi-devel] Re: [Freeipmi-users] bmc-autoconfig: drop for 0.5.0 rel


From: Al Chu
Subject: [Freeipmi-devel] Re: [Freeipmi-users] bmc-autoconfig: drop for 0.5.0 release?
Date: Wed, 05 Sep 2007 15:44:21 -0700

Hey Peter,

> Perhaps what is needed is a wiki or some such where real world
> problem/solutions examples can accumulate?

For FreeIPMI 0.5.0, I've been adding a lot of additional information to
the FreeIPMI manpages.  For example:

---
GENERAL USE
       Most users of bmc-config will want to:

       A)  Run  bmc-config  with --checkout to get a copy of the current
BMC configuration and store it in a file. The standard output
       can be redirected to a file or a file can be specified with the
--filename option.

       B) Edit the configuration file with an editor. See bmc-
config.conf(5) for information on what the fields in  the  configuration
       file mean.

       C) Commit the configuration back to the BMC using the --commit
option and specifying the configuration file with the --filename
       option.

       For users with large clusters or sets of nodes, you may wish to
use the same configuration file for all nodes. The one  problem
       with  this  is  that the IP address and MAC address will be
different on each node in your cluster and thus can't be configured
       through the same config file. The IP address and MAC address in
your config file may be overwritten on the command  line  using
       --key-pair option. The following example could be used in a
script to configure each node in a cluster with the same BMC config
       file. The script only needs to determine the correct IP address
and MAC address to use.

       # bmc-config --commit -k Lan_Conf:Ip_Address=$MY_IP -k
Lan_Conf:Mac_Address=$MY_MAC -f my_bmc.conf
---

Hopefully text like that will get users going where-as it may have been
more confusing before.  I also have pointers to bmc-config.conf(5)
(whereas there wasn't a pointer before, so most would not have seen the
manpage).  There are also trouble-shooting sections for generic issues.
I don't currently have a bmc-config specific trouble-shooting section.
Do you think that would be useful?  What kind of stuff do you think
should be in it?

Another thought I've had is adding additional sectional "header
comments" into the sections bmc-config checkout.  So for example:

#
# Section LAN_Conf
#
# This section is for configuring blah blah blah.  For most
# systems you want to configure blah blah blah.
Section LAN_Conf
   ...
EndSection

So that might give the user additional help in setting up their system.
Do you think that would really help?

> Where can a user find out if their machine supports SOL, and what
> could they do with it if it did? (rhetorical question, but was real 
> for me once.)

Hmmm.  That's a far harder question.  Outside of a specific list, I'm
not really sure what could be done.  Some vendors in the past have said
they support IPMI when they don't. :-)

Al

On Wed, 2007-09-05 at 15:24 -0700, Peter Broadwell wrote:
> This is understandable.
> Loosing the commented template is sad however.
> 
> Perhaps what is needed is a wiki or some such where real world
> problem/solutions examples can accumulate?
> 
> Where can a user find out if their machine supports SOL, and what
> could they do with it if it did?
> (rhetorical question, but was real for me once.)
> 
> ;;peter
> 
> 
> Al Chu wrote:
> > I had begun working on a template to store in the docs directory, with
> > comments throughout the file to inform the user of what they should
> > configure on their own.
> > 
> > However, with so many different BMCs and vendor implementations out
> > there, a substantial portion of the default template will fail for
> > different users and different hardware.  I think that will simply cause
> > confusion.  For example, a user may believe they have SOL configured
> > properly when their machine may not support SOL.
> > 
> > I'm more inclined to let the user run --checkout on their own, since it
> > will allow the user to configure exactly what is available for their
> > machine.  It is the model that LLNL and most users of FreeIPMI (that
> > I've spoken to) follow.
> > 
> > So for the time being, I've removed bmc-autoconfig.  If it can be
> > revamped to handle SOL, varying number of users, passwords, varying BMC
> > implementations, etc.  I think we can add it back in.
> > 
> > Al
> > 
> > On Thu, 2007-08-30 at 10:14 -0700, Al Chu wrote:
> >> I have been working with a user on a BMC config issue with their
> >> machine.  I'm now disinclined to support the committing of a default
> >> template.  
> >>
> >> 1) Many different machines support different configuration options.
> >> Included in this are: ipmi 1.5 only options vs ipmi 2.0 options vs.
> >> optionally supported options vs. newer errata options vs. flat out
> >> unsupported options.  So do we support the full template (so most
> >> options will fail by default) or do we support a minimal template (most
> >> options aren't listed).
> >>
> >> 2) Due to the IP address and MAC address being required for modification
> >> (and likely subnet + gateway too), at minimum, the user must edit the
> >> template anyways, we cannot create a default template that will work
> >> without modification.
> >>
> >> I think the better idea is to store a template in the docs location and
> >> mention it in the bmc-config manpage.  I have also written into the bmc-
> >> config manpage some general use instructions, so they know they should
> >> run --checkout to create a config template first.
> >>
> >> Any thoughts?
> >>
> >> Al
> >>
> >> On Fri, 2007-08-17 at 11:18 -0700, Anand Babu Periasamy wrote:
> >>> Hi Al,
> >>> I am thinking, if we produce $prefix/etc/freeipmi/bmc-config.conf with
> >>> fully documented options and default values, bmc-autoconfig's goal can
> >>> be achieved. Additionally it can be used for automation too.
> >>> bmc-config will use this config file if none is specified through the
> >>> command line argument. Then we can get rid of bmc-autoconfig. What do
> >>> you think?
> >>>
> >>> Al Chu writes:
> >>>
> >>>> I just thought of this.  We could also distribute a common template file
> >>>> as part of FreeIPMI and install it in the docs dir?  I guess my semi-
> >>>> argument against this is the fact that we've (practically) already
> >>>> distributed a template file with the bmc-config.conf(5) manpage.  So
> >>>> would there be a need?
> >>>>
> >>>> What are people's thoughts?
> >>>>
> >>>> Al
> >>>>
> >>>> On Tue, 2007-08-14 at 11:01 -0700, Peter Broadwell wrote:
> >>>>> I have need to configure many machines at the same time and if the 
> >>>>> templateing
> >>>>> file was documented this tool might become the one of choice for such 
> >>>>> uses.
> >>>>>
> >>>>> ;;peter
> >>>>>
> >>>>>
> >>>>> Anand Babu Periasamy wrote:
> >>>>>> Hi Al,
> >>>>>> * It is still maintained.
> >>>>>> * BMC-Autoconfig is not a GUI wizard for bmc-config. It is supposed to
> >>>>>>  ask minimum questions from the user and automatically configure the
> >>>>>>  BMC with known defaults. It is intended for users without any
> >>>>>>  knowledge of IPMI to quickly get a basic working setup.
> >>>>>>
> >>>>>> * It does enable LAN and configure NULL, admin, operator and ipmiuser
> >>>>>>  accounts. See the template file, you will get an idea what all it
> >>>>>>  configures.
> >>>>>> If you have suggestions to improve, let us know?
> >>>>>>
> >>>>>> Albert Chu writes:
> >>>>>>
> >>>>>>> I'm thinking of dropping this from FreeIPMI:
> >>>>>>>
> >>>>>>> A) It doesn't seem to be maintained by the original authors.
> >>>>>>>
> >>>>>>> B) It apparenly only configures 3 fields of the BMC.  No users, lan
> >>>>>>> enabling, etc.
> >>>>>>>
> >>>>>>> I don't really see the use anymore.  Any comments?  Anyone out there
> >>>>>>> using
> >>>>>>> this?
> >>>>>>>
> >>>>>>> Al
> >>>>>>>
> >>>>>>> -- 
> >>>> -- 
> >>>> Albert Chu
> >>>> address@hidden
> >>>> 925-422-5311
> >>>> Computer Scientist
> >>>> High Performance Systems Division
> >>>> Lawrence Livermore National Laboratory
-- 
Albert Chu
address@hidden
925-422-5311
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory




reply via email to

[Prev in Thread] Current Thread [Next in Thread]