Editor's Note: Minutes received 11/14/92
INTERIM_MEETING_REPORT_
Reported by Steven Waldbusser/CMU
Minutes of the Host Resources MIB Working Group (HOSTMIB)
Minutes of the Host Resources MIB Working Group held October 12-13, 1992
at Carnegie Mellon University, Pittsburgh PA
Agenda
o Introduction (Waldbusser)
- Local Arrangements
- Attendance
- Agenda
o Overview of Changes (Grillo, Waldbusser)
o Questions about current draft (Discussion)
o Proposals
o Summary
Overview of Changes from Last Draft
o New DateAndTime syntax (more changes in the works)
o Processor is now a device
o Storage allocation units
o New device types
o Nuke device location (as per mailing list)
o Add device status
o Add device errors
o Nuke video table (as per mailing list)
o Rearrange disk/partition/file system
o Full/Partial backup date
o Printer table
o SW type (OS, device driver,application)
Resolutions to the Proposed Topics
o Storage -> Device - explain logical vs. physical
There were requests to clarify the differences between the
hrStorage table and why it isn't in the hrDevice table. The
hrStorage table is defined as a table of logical storage entries as
opposed to the hrDevice table, which describes physical devices.
It was agreed that additional dialogue in the MIB was needed in
order to clarify the use of the hrStorage table.
1
o ProductID - Guidance to Industry, Should CMU Bootstrap?
It was agreed that ProductIDs are a good idea. There was
discussion as to how to encourage adoption of ProductIDs throughout
the industry. There was a proposal to `seed' a list in CMU address
space with definitions for popular products.
o Logging
It was decided that logging of system `events' was a useful
function, but was outside of the scope of the host resources MIB
effort.
o Nuke hrDiskStorageType
This enumerated list was dropped. This same information may be
determined from the hrDeviceType.
o ReadWrite/readOnly in disk/FS Entries
There was confusion about what readWrite/readOnly meant in the disk
and file system entries. This was clarified (see new MIB for
clarification).
o Asset Management - Serial #, ...
It was suggested that an object should exist for serial numbers of
devices (and possibly S/W). The Group agreed that, instead, the
hrDeviceDescr and hrSwInstalledName could contain the serial
number.
o Printer Status - deviceStatus relationship
See `general printer discussion' below.
o Nuke outOfPaper, Replace with interventionRequired
See `general printer discussion' below.
o Additional Variables for Network Table
A proposal was made for some additional variables to be added to
the network table. After some discussion, it was determined that
the proposed variables were PC NOS specific or belonged in a
networking MIB. The proposal was then dropped.
o Networked Disks
It was agreed that it would be important to identify those file
systems that were remotely mounted, but that the host MIB should
not support remote troubleshooting of these file systems (it should
contain enough information for a manager to find the file server
and communicate directly with it). A new object,
hrFSRemoteMountPoint was added to the hrFSTable to provide this
linkage.
2
o References to (MIB-II) System Group
The host resources MIB requires the use of certain MIB-II System
Group variables. There was concern about the discussions going on
in the SNMPv2 Working Group about the unclear semantics of the
sysDescr object. It was agreed that we would monitor these
activities and not duplicate them here, but if at some later date
this issue is not resolved we may need to add this functionality
into this MIB.
o General Printer Discussion
Considerable discussion occurred on how to best define printers in
this MIB. It was agreed that this MIB was not the place to have
total printer knowledge, but that generic printer state information
was needed.
Some problems with the October draft were that the printerStatus
variable was overloaded with a couple of orthogonal issues. The
usefulness of the tonerStatus variable did not warrant its own
variable in the MIB.
The printer group was redesigned and is included in the November
draft ofi the MIB. The relationship between the deviceStatus object
and the printerStatus object was refined and all the common error
states were moved into a bit string.
o ``Other''
It was agreed that an enumeration for ``other'' would be included
wherever appropriate throughout the MIB.
o Re: # of Users; What is a User?
A discussion took place over the definition of a user. There were
two possibilities: a human, or a distinct user session (possibly
many per human). The latter definition was chosen because it was
1) easier to calculate and, 2) correlated better with system
resource use.
o Processor Load
Due to the difficulty in calculating and interpreting processor
load using the semantics defined in the October draft, it was
decided that hrProcessorLoad was an average percentage of load
between 0 and 100% computed over the last minute.
o Parity Errors
A proposal was made to add memory/disk parity errors. After
discussion, we decided to use the existing deviceErrors variable
for these errors.
o 5, 15 Minute Averages
It was suggested that averages should exist over 5 and 15 minute
intervals. This idea was dropped as it was possible to determine
3
these averages from the 1 minute average currently in the MIB.
o SW Run CPU Used
There was a proposal to add this object and general agreement that
it was useful, but there was also a concern about the costs of
implementation on some systems. The resolution was to create the
object in a new group, the optional Running Software Performance
group. In addition, the hrSWRunSize object was moved into this new
group.
o SW Installed Path
The Group agreed that the hrSWInstalledPath object was not very
usable, o so it was deleted.
o SW Run Index/Status
There was some concern that with the ability to kill processes, the
hrSWRunIndex might not be ``unique enough'' and that the wrong
process could be killed. The resolution was to suggest that the
agent use the operating system's native process identification
scheme (i.e., the process id).
o SNMPv2 Format
It was agreed that the working group would not produce a document
with the new macros defined in SNMPv2, but that the Working Group
Chair would keep a (separate) working copy in that format to use as
an aid in rigorously identifying conformance and grouping issues
that should be addressed by the working group and described in
comments in the MIB.
Open Issues
o Networked Printer Identification
A proposal was made that networked printers and their addresses be
identified for all such printers ``connected'' to the system. This
would be useful for remote troubleshooting of printing problems.
The debate was over what ``connected'' meant in an environment when
network printing protocols typically require no state to be saved
on either end. In addition, many systems don't have any idea what
printers their applications are accessing, especially in multiuser
systems, where there is no common place to look for this
information.
No consensus was reached on whether or not to include this
functionality.
o Queue Monitoring
A proposal was made to add a table for monitoring queues (print,
mail, ...). The three options discussed were:
4
1. Don't add any support for queue monitoring.
2. Leverage off of similarities between queues and allocated
memory by monitoring queues with the hrStorageTable.
3. Add a table similar to the following:
HrQueueEntry ::= SEQUENCE {
hrQueueIndex INTEGER,
hrQueueDescription DisplayString,
hrQueueLength INTEGER,
hrQueueServicedItems Counter,
hrQueueInsertFailures Counter
}
Attendees
Amatzia Ben-Artzi amatzia@netmanage.com
Steve Bush sfb@ncoast.org
Raymond Edgerton bellmf!edgerton@uwm.edu
Michael Erlinger mike@lexcel.com
Pete Grillo pl0143@mail.psi.net
Todd Kulick TK24@andrew.cmu.edu
Ed Reeder ereeder@vnet.ibm.com
Jon Saperia saperia@tcpjon.ogo.dec.com
Steven Waldbusser waldbusser@andrew.cmu.edu
Walter Wong wcw@cmu.edu
5