How to run simple single command on multiple Linux servers?

Shell Programming and Scripting
1

Hi All,

How can i run a single command on multiple servers with or without giving credentials.

I have a file(servers.txt) which has got list of servers and i want to run a command

lsb_release -dr

on all these servers and get output of those servers against each server.

I tried below code but it didn't work

tmpdir=${TMPDIR:-/tmp}/pssh.$$
mkdir -p $tmpdir
count=0
while IFS= read -r userhost; do
    ssh -n -o BatchMode=yes ${userhost} 'lsb_release -dr' > ${tmpdir}/${userhost} 2>&1 &
    count=`expr $count + 1`
done < servers.txt
while [ $count -gt 0 ]; do
    wait $pids
    count=`expr $count - 1`
done
echo "Output is in $tmpdir"

output of $ lsb_release -dr

Description:    Red Hat Enterprise Linux Server release 6.4 (Santiago)
Release:        6.4

is there any single line command which can give me the output as below

gbahelkm72.trp.expre.rmr   Release:        6.4
rtsahelkm98.trp.expre.rmr   Release:        6.4
2

Hi,

Here is some code that you should be able to easily adapt, this was one of the subroutines used to collect data from over 600 servers prior to a DCM - just adapt and substitute your own bits.

for HOST in `cat ${CONFDIR}/svr-access.conf | grep -v '^#'`
do
        OUTFILE=details-${HOST}.csv
        echo "checking ${HOST}"
        MSG "checking ${HOST}"
        HOSTNAME=`ssh ${HOST} "uname -n"`
        OS=`ssh ${HOST} "uname"`
        # ssh_get_os_details ${HOST} ${OS}
	MSG "Found ${HOSTNAME} running ${OS}"
MSG "JOB STEP 2"
        if [ ${OS} = "AIX" ]
        then
MSG "JOB STEP 3"
                if [ `echo ${OSVER} | cut -c1` -ne 4 ]
                then
                        ssh_get_aix56_getdetails ${HOST} > ${DATADIR}/${OUTFILE}
                fi
	elif [ ${OS} = "SunOS" ]
	then
MSG "JOB STEP 4"
	${FUNCDIR}/ssh_solaris_getdetails ${HOST} > ${DATADIR}/${OUTFILE}
        fi
done

Regards

Gull04

---------- Post updated at 12:51 PM ---------- Previous update was at 12:35 PM ----------

Hi,

Just for some more information, here is one of the more simple routines that are called in the ssh calling script.

############################################################################################
#
# Get the Memory info.....
#
############################################################################################

get_mem()
{
GETMEM=`ssh ${SSHHOST} "/usr/sbin/prtconf 2>&1 | grep -v devinfo | grep 'Memory size' | /usr/xpg4/bin/awk '{ print $1 }'"`
MEM=`echo ${GETMEM} | awk '{ print $3 }'`
GBMEM=`echo ${MEM} /1024 | bc`
MSG "${SSHHOST} Memory Check Executed."
}

Regards

Gull04

3

Note the awk path in the above example is only valid for Solaris. Not Linux.

/usr/bin/awk should be okay for Linux. Note: this likely is a symlink to /usr/bin/gawk which is the GNU version of awk.

ls -l /usr/bin/awk

will show you what is going on

4

What's wrong? Do you get an error message?
I have omitted some stuff where I don't see a sense, and have prepended a hostname.

tmpdir=${TMPDIR:-/tmp}/pssh.$$
mkdir -p $tmpdir
echo "Output is in $tmpdir"
while read userhost
do
    (
    printf "%-29s " "$userhost"
    ssh -nx -o BatchMode=yes "$userhost" 'lsb_release -sd'
    ) </dev/null >"$tmpdir/$userhost" 2>&1 &
done < servers.txt
# wait until all bg jobs are done
wait
5

It is giving me the same host files which are there in servers.txt

6

Yes, I have assumed they are hostnames from servers.txt.
You did not say where gbahelkm72.trp.expre.rmr and rtsahelkm98.trp.expre.rmr come from.?

7

These are from servers.txt file.. this file has got hostnames

8

I'm not at all sure that I understand what you're trying to do, but note that the each of the background subshells started by:

    (
    printf "%-29s " "$userhost"
    ssh -nx -o BatchMode=yes "$userhost" 'lsb_release -sd'
    ) </dev/null >"$tmpdir/$userhost" 2>&1 &

writes to a different file for each host. If you want all of the output in a single file and you want to trim off the first part of the lsb_release output you need something more like:

    (
    release=$(ssh -nx -o BatchMode=yes "$userhost" 'lsb_release -sd')
    printf "%-29s %s\n" "$userhost" "Rel${release##*Rel}"
    ) </dev/null >>"$tmpdir/alluserhosts" 2>&1 &

clearing the output file before you start your loop with:

>"$tmpdir/alluserhosts"

unless you want the output in that file to grow each time you invoke your script.

Since I don't have access to a system with lsb_release installed, the above is totally untested. But, in theory, it should come close to what you seem to want.

9

Hmm I think asynchronous background jobs MUST go to separate logs.
Otherwise they might overwrite each other, and the shared log file becomes a mess.

10

Note that in the code I suggested in post #8, the output to the single log file is produced by a single invocation of the printf utility that uses an append redirection operator ( >> ) instead of the overwrite redirection operator ( > ) that was in the original code. Although not required by the standards, I would expect the output of that invocation of printf to be produced by a single write() system call, which, when writing the amount of data shown in the examples to a regular file, should be an atomic operation. Using the append redirection should prevent the concurrent background processes from overwriting each other's output and the single printf instead of multiple echo s should keep the output from the various background processes from being intermixed.