talk user
talk user@host

See who is logged in and what tty.

who -lu
logins -a

Who is an active login, who is idle.

finger -i

Print an error message via buffer

fold | lp
STUFF TEXT
^D

Save an error message via buffer

cat > somefile
STUFF TEXT
^D

Share the output with another user.

CMD | tee /dev/pts/THEIR_TTY CMD > /dev/pts/THEIR_TTY

Share the output with another user - Continuous.

script | tee /dev/pts/THEIR_TTY

Filesystems and Files

ls -ld /some/dir

To see local mounted filesystems.

df -kl

To see where a directory is really mounted from.
Local will start with /dev, NFS will be node:, memory will be text and /proc

df -k /some/path
cd /some/path; df -k .

To view the current filesystem size by directory.

du -sk * | sort -nr | head

Some filesystem fills up, but du -sk does not help.

find . -mtime -1 #find files modified with a day
find . -mtime -1 -ls OR
find . -size +500 #find files over 500 blocks in size find . -size +500 -ls

A place to look if root fills up, but nothing else looks suspect.

find /dev -type f -ls # look for large file (bogus tar device/filename)

Normal OS logging dirs.

/var/adm
• /var/log
• /var/spool
• /tmp #memory only - shares with swap
• /var/tmp
• /var/crash/HOSTNAME

find . -type f -mtime +3 exec compress {} \;

We have a wide range of Sun hardware using Sparc processors. Most nodes are multiprocessor. To list number and speed of processors: psrinfo -v.

Our nodes are reached via terminal consoles made by Perle. Only one user can be on a console at a time. There are logged into via telnet with a terminal type of vt100. Messages written to the console are lost if the console is not active (logged into), but we do not normally leave them logged in because we want to access consoles from anywhere.

If a node crashes, and does not reboot, we want to catch register values on the console (/notes/crashed_system).

Ops can now do a normal reboot of a node when needed by permission of the Unix Oncall Admin using sudo. Steps as follows.

NOTE: init states are as follows:

1. 0 - halt
2. 1 - single-user
3. S - single-user
4. 5 - power-off (must cycle power switch to turn it back on)
5. 6 - reboot

1. /scripts/console HOSTNAME #user, password, Return
2. Login as yourself.
3. /NFS-SHARE/admin/bin/sudo /usr/sbin/shutdown -g30 -y -i6

If a node is hung. Work closely with the Unix SA oncall in this situation.

1. /scripts/console HOSTNAME #user, password, Return
2. ^] #control-]
3. send break
4. boot

/NFS-SHARE/admin/bin/suntype -v

To see the hardware layout and hardware errors (in addition to /var/adm/messages).

/usr/platform/sun4u/sbin/prtdiag -v

Messages and Errorlogs files.

/var/adm/messages
/var/log/syslog

uptime
who -b #no loading stats

Memory usage and processor usage.

vmstat 2
vmstat -S 2
vmstat -s

Disk I/O usage.

iostat -xtnc 2
iostat -xnp 2
iostat -DMnpx 2
iostat -Een #disk types *** great command - remember ***

CPU usage.

mpstat 2

System usage - historical.
There are many options - do a 'man sar'. Similar to Patrol.

sar -A # list all available information

See process usage.

/NFS-SHARE/admin/bin/sps cpu #by percent cpu
/NFS-SHARE/admin/bin/sps rss #by percent memory

See swap usage.

swap -l
swap -s

/etc/password
/etc/shadow
/etc/group
/etc/aliases

Commands to look at NIS information. NIS is on the internal network only - not on the DMZ

/usr/bin/ypmatch username passwd
/usr/bin/ypmatch username group
/usr/bin/ypcat passwd
/usr/bin/ypcat -k aliases
/usr/lib/sendmail -bv username

See user info.

finger username

See groups currently active.

groups

Processes

/usr/proc/bin/* (do a man on each one)
truss -aefp PID
/usr/openwin/bin/sdtprocess
ps -cafel OR /opt/admin/bin/sps (sort by CPU)
prstat -a (top like output in newer Solaris)

MEMORY
/usr/bin/prtmem
/usr/bin/memtool
/usr/proc/bin/pmap -x PID
/usr/openwin/bin/sdtprocess
ps -cafel OR /opt/admin/bin/sps (sort by RSS or SZ)

Sendmail logs its information to syslog.

The syslog file (/var/log/syslog) rolls over depending on size/time-period and have a numeric extension - higher the number, older the file. An "ls -l /var/log/syslog*" shows the date they were last modified (added to).

Sendmail processes
The count should be greater that 1. We like to see it around 30. Run this on mailserver or the backup mailhost.

mailserver: ps -ef | grep -c sendmail

See number of messages in queue.
Each mail messages takes approixmately 3 lines, so to get a count of jobs in the queue:

mailserver: ls /var/spool/mqueue/qf* | wc -l

Find a users activity.

mailserver: grep USERNAME syslog

Watch for a users activity.

mailserver: tail -f syslog | grep USERNAME # ^C when your are done

See OS errors (and some other mail messages not is syslog).

mailserver: tail -100 /var/adm/messages

Send a test email (you can have the "tail -f syslog" running too)

/usr/ucb/mail -v someuser@some.domain
>Subject: test
>testing
>.

Show aliases for username

mailserver: /usr/lib/sendmail -bv USERNAME

See what version of sendmail you are running (we hide the version on mailserver):

mconnect mailserver
^D exits
OR
telnet mailserver 25
^]

View stats - (man mailstats to learn about colums/entries)

mailserver: /bin/mailstats
OR
mailserver: cd /var/log/mailstat
then look for the date you want (mailstats saved to a file)

We have an internal DNS and an external DNS. Internal is done by XXX, external by XXX for the DMZ and our ISP (XXX) slaves our domain.

Nslookup: to look up a hosts IP address

nslookup hostname
OR
nslookup
> hostname
> ^D

Nslookup: to look up a domain's address

nslookup
> ls mydomain.com > /tmp/myfile
> ^D
more /tmp/myfile

lpstat -a

print out outstanding print jobs

lpstat -o [printername]

To find out the printer name a queue uses on a specific host

lpstat -a
find the queue name you want
IF you are using Sun/Unix printer software:

cd /etc/lp/printers;ls
grep dest= queue_name/configuration
grep dest= myqueue/configuration
#Use the dest=VALUE for printer DNS name
Options: protocol=bsd,dest=myqueue

IF you are using HP printer software:

cd /etc/lp/interfaces;ls
grep PERIPH= queuename
grep PERIPH= project
#Use the dest=VALUE for printer DNS name
PERIPH=project

The output of the grep will give the DNS name of the printer

Sudo commands for printer management, use with direction of Unix admins

/NFS-SHARE/admin/bin/sudo cancel [printer, request-id, and/or user]
/NFS-SHARE/admin/bin/sudo enable queuename
/NFS-SHARE/admin/bin/sudo disable queuename
/NFS-SHARE/admin/bin/sudo lpmove request-id queuename
/NFS-SHARE/admin/bin/sudo lpmove queuename-orig queuename-new

1. cscope : manage source tree (global searches, replaces, etc., by function, variable, ...)
2. d2h, h2d : change a decimal number to hex and back again
3. day2julian, julian2day : change a date to julian number and back again
4. get_service_name service-name : to see if a particular (exact) service name is on a host (calls srvnames)
5. get_all_service_name [pattern] : to see if either a patten (part of a name) is on a server OR to see all the service names on a server (no argument is given for this) (calls srvnames)
6. grepc : grep that lists lines before and after pattern match
7. grepr : recursive grep that also hanldes binary files
8. hex2ip, ip2hex : change an IP address to hex and back again
9. suntype : list the stats about the host you are on. Use the "-v" for most o utput.
10. sps : a sorted (static) ps
11. srvnames : has many options, list service names, search for a service name, add or delete a service name from a host. Can also mount/umount disks for specf ic service name. Some actions require root privilege.
12. what_date : given an offset, compute a date from current date
13. xdiff : graphical diff - very nice

Sh & Ksh Info

How to leave a job running when you logout or you loose your window connection:

(For Bourne (sh) and Korn (ksh) only. C-Shell (csh) - use the & and it will alr eady disconnect from startup terminal.)

cmdprompt> nohup someprocess &

PC & Unix Interoperability Tools