Murali Suriar's questions

What are VLANs? What problems do they solve?

I'm helping a friend learn basic networking, as he's just become the sole sysadmin at a small company. I've been pointing him at various questions/answers on Serverfault relating to various networking topics, and noticed a gap - there doesn't appear to be an answer which explains from first principles what VLANs are. In the spirit of How does Subnetting Work, I thought it would be useful to have a question with a canonical answer here.

Some potential topics to cover in an answer:

• What are VLANs?
• What problems were they intended to solve?
• How did things work before VLANs?
• How do VLANs relate to subnets?
• What are SVIs?
• What are trunk ports and access ports?
• What is VTP?

EDIT: to be clear, I already know how VLANs work - I just think that Serverfault should have an answer that covers these questions. Time permitting, I'll be submitting my own answer as well.

I have been using gnu-screen's 'logfile' command to log my session histories when making production changes. I like to log each change in a different directory; this means that whenever I start a new change, I need to execute the :logfile /path/to/change/directory/%t.log command again. I was wondering if there is a way to specify a logfile string when invoking screen?

The goal here is to be able to write a wrapper script, e.g. startchange [changenumber] which would then change to the correct change directory and configure screen to log any sessions in that directory.

The only solution I can think of at the moment is to have a script that makes a copy of my global .screenrc, replaces the logfile string, and invokes screen with the '-c' command line option. (screen doesn't appear to support reading multiple config files either.) This feels like a fairly heavyweight solution to what should be a fairly simple problem. Does anyone have any alternative approaches?

I am attempting to poll the temperature of an HP ProCurve 3500YL switch remotely using SNMP.

Looking at this MIB, it appears that the following OIDs:

hpCpuTemperature            1.3.6.1.4.1.11.2.3.7.11.17.7.1.1.1.6
hpPowerSupplyTemperature    1.3.6.1.4.1.11.2.3.7.11.17.7.1.1.1.7
hpChassisTemperature        1.3.6.1.4.1.11.2.3.7.11.17.7.1.1.1.8

Within the 'hpProcurveSysMib' should provide the data I need. However, whenever I attempt to access these OIDs, I receive the response:

SNMPv2-SMI::enterprises.11.2.3.7.11.17.7.1.1.1.6 = No Such Object available on this agent at this OID

Further investigation reveals that the switch in question does not appear to implement the parent hpProcurveSystem MIB:

SNMPv2-SMI::enterprises.11.2.3.7.11.17.7.1.1 = No Such Object available on this agent at this OID

Does anyone know of an alternative MIB implemented by the 3500 that will allow its temperature to be polled automatically?

I have a custom bash prompt that colours my username, hostname and current working directory. The colours display correctly, however when I type a command line that needs to wrap, the terminal does not start a new line, and writes over the beginning of my command. This occurs on RHEL 4 and Debian.

This question details a similar problem on OS X; it would seem that this is an issue with incorrectly terminated control sequences, however I can't see what could be causing this, as I am using $(tput) codes everywhere, rather than hand crafting my codes. Snippet below - is anyone able to advise what I am doing wrong?

# Control codes

COL_RST=$(tput sgr0)  # Remove all colour formatting
COL_BRIGHT=$(tput bold) # Emphasise

# Foreground
COL_G=$(tput setaf 2) # Green
COL_Y=$(tput setaf 3) # Yello
COL_B=$(tput setaf 4) # Blue

COL_USR=${COL_Y}${COL_BRIGHT}
COL_PWD=${COL_B}${COL_BRIGHT}
COL_HST=${COL_G}${COL_BRIGHT}

# Set prompt
PROMPT="${COL_USR}\u${COL_RST}@${COL_HST}\h${COL_RST}:${COL_PWD}\W${COL_RST}"

# Put it all together....
PS1="[$PROMPT]\$ "

export PS1

Does anyone have any experience of using Kerberos as an authentication mechanism for managing a Cisco IOS based network? This article seems to indicate it's possible, but my Kerberos knowledge is limited to that of a user on centrally managed UNIX/Linux systems. Before I spend hours trying to lab this up, I thought I would ask for advice here.

Specifically, given a functioning Kerberos authentication infrastructure, is it possible to configure Cisco devices to accept forwarded Kerberos credentials from a *NIX or Windows machine where I have already authenticated and have a valid Kerberos ticket? It would be really nice not to have to type my password every single time I log in to a device.

If this is possible, some further extensions:

1. Is it possible to use Kerberos for authentication, but to continue to use TACACS+ groups for authorisation?

2. Under what conditions will a device configured to use Kerberos authentication fall back to locally defined passwords?

3. What implications does using Kerberos have for performing RMAs/hardware replacements? (e.g., if using SSH only as a means of management, if a device is replaced, SSH keys must be regenerated manually and old known_hosts entries on management stations must be removed, etc)

4. When using Kerberos authentication, will a user still be prompted for a password when elevating from EXEC mode to privileged EXEC (enable) mode?

If I have the following entries in a hosts file:

192.168.100.1    bugs
192.168.100.2    daffy.example.com
192.168.100.3    elmer.example.com.

Will IP->name resolution attempts by local utilies (I assume using 'gethostbyaddr' or the Windows equivalent) honour these entries? Is this behaviour configurable? How does it vary between operating systems? Does it matter whether the 'hosts' file entries are fully qualified or not?

EDIT: In response to Russell, my test Linux system is running RHEL 4. My /etc/nsswitch.conf contains the following 'hosts' line:

hosts:      files dns nis

If I ping any of my hosts by name (e.g. bugs, daffy), the forward resolution works correctly. If I traceroute any of them by IP address, the reverse lookup functions as expected. However, if I ping them by IP, ping doesn't appear to resolve their host names. My understanding was that Linux ping would always attempt to resolve IPs to names unless instructed otherwise. Why would traceroute be able to handle reverse lookups in hosts files, but ping not?

I discovered through this answer that I've been setting terminal colours in my prompt in a naive way for years.

I've now modified my .bashrc to use 'tput' commands to colourise various elements of my prompt. The one remaining escape code I have is this:

TITLEBAR="\[\033]2; PROD - \u@\h:\w \007\]"

Does anyone know if there is a tput/terminfo attribute I can use to set the title bar of my terminal window, so that I can eliminate that escape?

Has anyone configured IIS 6.0 loadbalancing on win 2003 servers with ASP.NET app using InProc Sessions. How can this be achieved?

On a Solaris 8 server, how do I find out what process is listening on a given port? The Windows implementation of 'netstat' offers the -o and -b switches, which show the owning process ID and name respectively. Is there an equivalent under Solaris?