I have a file forwarding system where a bunch of files are downloaded to a directory, de-multiplexed and copied to individual machines.
The files are forwarded when they are received by the master server. And files normally arrive in bursts. (Auth by ssh keys)
This script creates the sftp session, and uses a pipe to watch the head of a fifo pipe.
HOST=$1
pipe=/tmp/pipes/${HOST%%.*}
ps aux | grep -v grep | grep sftp | grep "user#$HOST" > /dev/null
if [[ $? == 0 ]]; then
echo "FTP is Running on this Server"
exit
else
pid=`ps aux | grep -v grep | grep tail | tr -s ' ' | grep $pipe`
[[ $? == 0 ]] && kill -KILL `echo $pid | cut -f2 -d' '`
fi
if [[ ! -p $pipe ]]; then
mkfifo $pipe
fi
tail -n +1 -f $pipe | sftp -o 'ServerAliveInterval 60' user#$HOST > /dev/null &
echo cd /tmp/data >>$pipe #Sends Command to Host
echo "Started FTP to $HOST"
Update: I ended up changing the cleanup code to use "ps aux" to see if an ftp session is running, and subsequently if the tail -f is still running. Grep by user#host and the name of the pipe respectively. This is done when the script is called, and the script is called whenever I try to upload a file.
IE:
FILENAME=`basename $1`
function transfer {
echo cd /apps/data >> $2 # For Safety
echo put $1 .$FILENAME >> $2
echo rename .$FILENAME $FILENAME >> $2
echo chmod 0666 $FILENAME >> $2
}
./ftp.sh host
[ -p $pipedir/host ] && transfer $1 $pipedir/host
Files received on the master server are caught by Incron which writes a put command and the available file's location to the fifo pipe, to be sent by sftp (rename is also preformed).
My question is, is this safe? Could this crash on ftp errors/events. Not really worried about login errors.
The goal is to reduce the number of ftp logins. Single Session/Minute(or more) intervals.
And allow files to be forwarded as they're received. Dynamic Commands.
I'd prefer to use standard ubuntu libraries, if possible.
EDIT: After testing and working through some issues the server simply runs with
[[ -p $pipe ]] && echo FTP is Running on this Server
ln -s $pipe $lock &> /dev/null || (echo FTP is Running on this Server && exit)
[[ ! -p $pipe ]] && mkfifo $pipe
( tail -n +1 -F $pipe & echo $! > $pipe.pid ) | tee >
( sed "/tail:/ q" >/dev/null && kill $(cat $pipe.pid) |& rm -f $pipe >/dev/null; )
| sftp -i ~/.ssh/$HOST.rsa -oServerAliveInterval=60 user#$HOST &
rm -f $lock
Its rather simple but works nicely.
you might be intrested in setting up a more simpler(and robust) syncronization infrastructure:
if a given host is not connected when a file arrives...it never recieves it (if i understand correctly your code)
i would do something like
rsync -a -e ssh user#host:/apps/data pathToLocalDataStore
on the client machines either periodically or by event...rsync is intelligently syncronizes the files by their timestamp and size (-a contains -t)
the event would be some process termination like:
client does(configure private key usage in ~/.ssh/config for host):
#!/bin/bash
while :;do
ssh user#host /srv/bin/sleepListener 600
rsync -a -e ssh user#host:/apps/data pathToLocalDataStore
done
on the server
/srv/bin/sleepListener is a symbolic link to /bin/sleep
server after recieving new file:
killall sleepListener
note: every 10 minutes a full check is performed...if nodes go offline/online it doesn't matter...
Related
Im trying to check connections for a list of servers. I want to loop through the list, check if a connection works and if yes, do some stuff, if not, echo out a problem message.
My problem is:
the script stops at the first node without echoing the $?.
So, whats wrong with my for-loop?
These vars are included from a config file:
$nodes is a list of server IPs like 1.1.1.1,2.2.2.2,10.10.10.10
$user is one string
for node in $(echo $nodes | sed "s/,/ /g")
do
echo "Checking Node: $node"
ssh -q -o ConnectTimeout=3 $user#$node echo ok
echo $?
if [[ $? != 0 ]]
then
echo "Problem in logging into $node"
else
# do some stuff here
fi
done
EDIT #1:
for node in $(echo $nodes | sed "s/,/ /g")
do
echo "Checking Node: $node"
ssh -q -t -o ConnectTimeout=3 $user#$node "echo ok"
retcode=$?
echo $retcode
if [[ "$retcode" -ne 0 ]]
then
echo "Problem in logging into $node"
else
echo "OK"
fi
done
It is because ssh first asks you to validate The authority of the host and If you accept the authority it will ask for password. That is why your command does not return to shell and waits for input.
If your intention is just validating ssh connection, then you may consider to use
telnet <your_host> <port> < /dev/null
But if your intend is to run some commands you need a trust relationship between hosts. In that case you can use:
Execute this commands:
ssh-keygen -t rsa
then
ssh-copy-id -i root#ip_address
Now you can connect with
ssh <user>#<host>
Furher information
You can add -tto make virtual terminal and add quotes on command:
ssh -q -t -o ConnectTimeout=3 ${user}#${node} "echo ok"
Also use -ne instead of != which is for compare strings
if [[ "$?" -ne 0 ]]
Also echo $? mess the return code. You should use something like:
ssh -q -t -o ConnectTimeout=3 ${user}#${node} "echo ok"
retcode=$?
echo $retcode
if [[ "$retcode" -ne 0 ]]
You can rewrite ssh command like this to avoid problems with ssh host keys
ssh -q -t -o StrictHostKeyChecking=no -o ConnectTimeout=3 ${user}#${node} "echo ok"
The script monitors incoming HTTP messages and forwards them to a monitoring application called zabbix, It works fine, however after about 1-2 days it stops working. Heres what I know so far:
Using pgrep i see the script is still running
the logfile file gets updated properly (first command of script)
The FIFO pipe seems to be working
The problem must be somewhere in WHILE loop or tail command.
Im new at scripting so maybe someone can spot the problem right away?
#!/bin/bash
tcpflow -p -c -i enp2s0 port 80 | grep --line-buffered -oE 'boo.php.* HTTP/1.[01]' >> /usr/local/bin/logfile &
pipe=/tmp/fifopipe
trap "rm -f $pipe" EXIT
if [[ ! -p $pipe ]]; then
mkfifo $pipe
fi
tail -n0 -F /usr/local/bin/logfile > /tmp/fifopipe &
while true
do
if read line <$pipe; then
unset sn
for ((c=1; c<=3; c++)) # c is no of max parameters x 2 + 1
do
URL="$(echo $line | awk -F'[ =&?]' '{print $'$c'}')"
if [[ "$URL" == 'sn' ]]; then
((c++))
sn="$(echo $line | awk -F'[ =&?]' '{print $'$c'}')"
fi
done
if [[ "$sn" ]]; then
hosttype="US2G_"
host=$hosttype$sn
zabbix_sender -z nuc -s $host -k serial -o $sn -vv
fi
fi
done
You're inputting from the fifo incorrectly. By writing:
while true; do read line < $pipe ....; done
you are closing and reopening the fifo on each iteration of the loop. The first time you close it, the producer to the pipe (the tail -f) gets a SIGPIPE and dies. Change the structure to:
while true; do read line; ...; done < $pipe
Note that every process inside the loop now has the potential to inadvertently read from the pipe, so you'll probably want to explicitly close stdin for each.
#!/bin/bash
ip route add 10.105.8.100 via 192.168.1.100
date
cat /home/xxx/Documents/list.txt | while read output
do
ping="ping -c 3 -w 3 -q 'output'"
if $ping | grep -E "min/avg/max/mdev" > /dev/null; then
echo 'connection is ok'
else
echo "router $output is down"
then
cat /home/xxx/Documents/roots.txt | while read outputs
do
cd /home/xxx/Documents/routers
php rebootRouter.php "outputs" admin admin
done
fi
done
The other documents are:
lists.txt
10.105.8.100
roots.txt
192.168.1.100
when i run the script, the result is a reboot of the router am trying to ping. It doesn't ping.
Is there a problem with the bash script.??
If your files only contain a single line, there's no need for the while-loop, just use read:
read -r router_addr < /home/xxx/Documents/list.txt
# the grep is unnecessary, the return-code of the ping will be non-zero if the host is down
if ping -c 3 -w 3 -q "$router_addr" &> /dev/null; then
echo "connection to $router_addr is ok"
else
echo "router $router_addr is down"
read -r outputs < /home/xxx/Documents/roots.txt
cd /home/xxx/Documents/routers
php rebootRouter.php "$outputs" admin admin
fi
If your files contain multiple lines, you should redirect the file from the right-side of the while-loop:
while read -r output; do
...
done < /foo/bar/baz
Also make sure your files contain a newline at the end, or use the following pattern in your while-loops:
while read -r output || [[ -n $output ]]; do
...
done < /foo/bar/baz
where || [[ -n $output ]] is true even if the file doesn't end in a newline.
Note that the way you're checking for your routers status is somewhat brittle as even a single missed ping will force it to reboot (for example the checking computer returns from a sleep-state just as the script is running, the ping fails as the network is still down but the admin script succeeds as the network just comes up at that time).
I have the following which works perfectly.
#!/bin/bash
killall java
#program USB
make iris install.1 mib510,/dev/ttyUSB0
#listen serial port and write to file
java net.tinyos.tools.PrintfClient -comm serial#/dev/ttyUSB1:iris > foo.txt &
sleep 2
#if "Erase done" is printed to file, stop
if tail -f foo.txt | grep -n "Erase done" -q; then echo "Write ok";fi
killall java
But when I change my script to receive arguments below (sh test.sh USB0 USB1 foo.txt), it does not end. Although it writes the file, the process does not end
#!/bin/bash
killall java
#program USB
make iris install.1 mib510,/dev/tty$1
#listen serial port and write to file
java net.tinyos.tools.PrintfClient -comm serial#/dev/tty$2:iris > $3 &
sleep 2
#if "Erase done" is printed to file, stop
if tail -f $3 | grep -n "Erase done" -q; then echo "Write ok";fi
killall java
Am I doing something wrong?
It appears tail -f will quit when grep quits. So the problem might be with:
if tail -f $3 | grep -n "Erase done" -q; then echo "Write ok";fi
You can replace it with the following:
tail -f $3 | while read LOGLINE
do
[[ "${LOGLINE}" == *"Erase done"* ]] && echo "Write ok" && pkill -P $$ tail
done
I'm working on a simple bash script daemon that uses Unix domain sockets. I have a loop like this:
#!/bin/bash
while true
do
rm /var/run/mysock.sock
command=`nc -Ul /var/run/mysock.sock`
echo $command > /tmp/command
done
I'm echoing the command out to /tmp/command just for debugging purposes.
Is this the best way to do this?
Looks like I'm late to the party. Anyway, here is my suggestion I employ successfully for one-shot messages with response:
INPUT=$(mktemp -u)
mkfifo -m 600 "$INPUT"
OUTPUT=$(mktemp -u)
mkfifo -m 600 "$OUTPUT"
(cat "$INPUT" | nc -U "$SKT_PATH" > "$OUTPUT") &
NCPID=$!
exec 4>"$INPUT"
exec 5<"$OUTPUT"
echo "$POST_LINE" >&4
read -u 5 -r RESPONSE;
echo "Response: '$RESPONSE'"
Here I use two FIFOs to talk to nc (1) and fetch it's response.
You can use a single file also to use bidirectional.
mkfifo communicate_pipe
exec 3<> communicate_pipe
cat communicate_pipe - | python socket.py 127.0.0.1:8002 | while read line; do
cmd="./something.sh '${line}' > communicate_pipe";
eval $cmd;
fi;
done