I have systemd service, say xyzWarmup.service.
Here is the service file
[Unit]
Description=Xyz agent.
After=fooAfter.service
Before=fooBefore1.service
Before=fooBefore2.service
[Service]
# During boot the xyz.sh script reads input from /dev/console. If the user
# hits <ESC>, it will skip waiting for xyz and abc to startup.
Type=oneshot
StandardInput=tty
StandardOutput=tty
ExecStart=/usr/bin/xyz.sh start
RemainAfterExit=True
ExecStop=/usr/bin/xyz.sh stop
[Install]
WantedBy=multi-user.target
Following is the part of xyz.sh.
#! /bin/bash
#
### BEGIN INIT INFO
# Required-Stop: Post
### END INIT INFO
XYZ=/usr/bin/Xyz
prog="Xyz"
lockfile=/var/lock/subsys/$prog
msg="Completing initialization"
start() {
# Run wfw in background
ulimit -c 0
# wfw has a default timeout of 10 minutes - just pick a large value
wfw -t 3600 xyz abc >/dev/null 2>&1 &
PID=$!
# Display the message here after spawning wfw so Esc can work
echo -n $"$msg (press ESC to skip): "
while [ 1 ]; do
read -s -r -d "" -N 1 -t 0.2 CHAR || true
if [ "$CHAR" = $'\x1B' ]; then
kill -9 $PID 2>/dev/null
# fall through to wait for process to exit
fi
STATE="`ps -p $PID -o state=`"
if [ "$STATE" = "" ]; then
# has exited
wait $PID 2>/dev/null
if [ $? -eq 0 ]; then
echo "[ OK ]"
echo
exit 0
else
echo "[ FAILED ]"
echo "This is failure"
exit 1
fi
fi
done
}
When this script runs during boot I see the following message coming from the script
Completing initialization (press ESC to skip):
Updated:
This is the additional output which I see after the previous line
[[ OK ] Started Xyz agent.\n'
If you carefully see, there are 2 opening square brackets( '[' ), from this it looks like that systemd is overwriting the log messages. The first "[" comes from the initscript's "[ OK ]". Can somebody explain this better ?
I don't see "[ OK ]" or "[ FAILED ]" on my screen.
When I was using this script as initscript in Fedora14, I used to see these messages. Once, I have shifted to systemd. I have started seeing this issue.
systemd version is : systemd-201-2.fc18.9.i686 and systemd.default_standard_output=tty
Kindly help.
It looks to me that your issue here is that the script is never getting attached to the TTY. The output is showing up because you have that hard-coded to go to /dev/console in your script. With StandardInput=tty, systemd waits until the tty is available, and it's probably already in use. Your script is just sitting there not connected to input in the infinite loop. You could try StandardInput=tty-force, and I bet that will work, although I'm not sure what else that might break.
Personally, I think I might go back and rethink the approach entirely. It sounds like you want the boot to entirely block on this service, but let you skip by hitting escape. Maybe there's a better way?
Related
I do not have much experience with bash scripts, but i got the idea from the internet.
My bash script uses xprintidle to shutdown after the computer is in idle for some time.
I can run the script in terminal without any problem.
But when the /etc/systemd/system/poweroff.service is calling the script it gives the error in the systemctl status.
Jul 30 16:43:40 godo systemd[1]: Started autopoweroff.
Jul 30 16:43:42 godo bash[3107]: couldn't open display
Jul 30 16:43:42 godo bash[3107]: /usr/local/bin/poweroff.sh: line 5: [: : integer expression expected
Jul 30 16:43:42 godo bash[3107]: end
Here is the script:
#!/bin/bash
sleep 2
myidle=$(xprintidle)
myidletime=$((10000))
while [ "$myidle" -le "$myidletime" ]; do
echo $myidle
sleep 1
myidle=$(xprintidle)
done
#sudo shutdown -P now
#shutdown -P 5
echo "end"
And here is the service:
[Unit]
Description=autopoweroff
[Service]
ExecStart=/bin/bash /usr/local/bin/poweroff.sh
[Install]
WantedBy=multi-user.target
I hope you can help me and I do not waste your time with these beginner questions.
Thanks
When xprintidle does not have a display it print: "couldn't open display", you are trying to then compare this invalid value as an interger using "-le".
Since xprintidle returns exit code 1 when it does not have a display, you can use
set -e
at the start of your script to exit when an error occurs.
xprintidle - utility printing user's idle time in X
When your script runs in the systemd context, it has no X server, so xprintidle fails and output couldn't open display to stderr.
Your statement myidle=$(xprintidle) causes the myidle assignment to fail.
At this point you have to decide what you want to do when the X environment is not available.
A possiblity, is to have myidle with a default 0 value:
typeset -i myidle # Tells Bash it is an int and default to 0 if not assigned a value
myidle=$(xprintidle 2>/dev/null) || true # no error state generated
I think you need another way to get the idle value of the current currently running X session.
Here it is:
#!/bin/dash
sleep 2
# get the X DISPLAY of the first logged-in user with a X session
DISPLAY="$(
w --short --no-header \
| awk '{ if( match($3, ":") ) { print $3; exit; } }'
)"
export DISPLAY
myidletime=$((10000))
while myidle=$(xprintidle 2>/dev/null) && [ "$myidle" -le $myidletime ]; do
echo "$myidle"
sleep 1
done
#sudo shutdown -P now
#shutdown -P 5
echo "end"
on my raspberry pi (raspbian running) I would like to have the current desktop switched to desktop n#0 after 5 minutes of idle system (no mouse or keyboard action), through wmctrl -s 0 and xprintidle for idle time checking.
Please keep in mind I'm no expert...
I tried 2 different ways, none of them working and I was wondering which one of them is the best way to do have the job done:
bash script and crontab
I wrote a simple script which checks if xprintidle is greater than a previously set $IDLE_TIME, than it switches desktops (saved in /usr/local/bin/switchDesktop0OnIdle):
#!/bin/bash
# 5 minutes in ms
IDLE_TIME=$((5*60*1000))
# Sequence to execute when timeout triggers.
trigger_cmd() {
wmctrl -s 0
}
sleep_time=$IDLE_TIME
triggered=false
while sleep $(((sleep_time+999)/1000)); do
idle=$(xprintidle)
if [ $idle -ge $IDLE_TIME ]; then
if ! $triggered; then
trigger_cmd
triggered=true
sleep_time=$IDLE_TIME
fi
else
triggered=false
# Give 100 ms buffer to avoid frantic loops shortly before triggers.
sleep_time=$((IDLE_TIME-idle+100))
fi
done
script itself works.
Then I added it to crontab (crontab -e) for have it run every 6 minutes
*/6 * * * * * sudo /usr/local/bin/switchDesktop0OnIdle
not sure sudo is necessary or not.
Anyway It doesn't work: googling around I understood that crontab runs in its own environment with its own variables. Even though I don't remember how to access this environment (oops) I do remember that I get these 2 errors running the script in it (which correctly works in "normal" shell)
could not open display (is it important ?)
bla bla -ge error, unary operator expected or similar: basically xprintidle doesn't work in this environment a gives back an empty value
What am I missing ?
infinite-while bash script running as daemon
second method I tried to set up a script with an internal infinite-while checking if xprintidle is greater then 5 minutes. In this case desktop is switched (less elegant?). Saved also in /usr/local/bin/switchDesktop0OnIdle
#!/bin/bash
triggered=false
while :
do
if [ `xprintidle` -ge 300000 ]; then
if [ triggered == false ]
wmctrl -s 0
triggered = true
fi
else
triggered = false
fi
fi
done
again the script itself works.
I tried to create a daemon in /etc/init.d/switchDesktop0OnIdle (really not an expert here, modified an existing one)
#! /bin/sh
# /etc/init.d/switchDesktop0OnIdle
### BEGIN INIT INFO
# Provides: switchDesktop0OnIdle
# Required-Start: $all
# Required-Stop: $all
# Should-Start:
# Should-Stop:
# Default-Start: 2 3 4 5
# Default-Stop: 0 1 6
# Short-Description:
# Description:
### END INIT INFO
DAEMON=/usr/local/bin/switchDesktop0OnIdle
NAME=switchDesktop0OnIdle
test -x $DAEMON || exit 0
case "$1" in
start)
echo -n "Starting daemon: "
start-stop-daemon --start --exec $DAEMON
echo "switchDesktop0OnIdle."
;;
stop)
echo -n "Shutting down daemon:"
start-stop-daemon --stop --oknodo --retry 30 --exec $DAEMON
echo "switchDesktop0OnIdle."
;;
restart)
echo -n "Restarting daemon: "
start-stop-daemon --stop --oknodo --retry 30 --exec $DAEMON
start-stop-daemon --start --exec $DAEMON
echo "switchDesktop0OnIdle."
;;
*)
echo "Usage: $0 {start|stop|restart}"
exit 1
esac
exit 0
I set it up
sudo update-rc.d switchDesktop0OnIdle defaults
and
sudo service switchDesktop0OnIdle start
(necessary?)
...and nothing happens...
also I don't find the process with ps -ef | grep switchDesktop0OnIdle but it seems running with sudo service switchDesktop0OnIdle status
can anyone please help?
thank you
Giuseppe
As you suspected, the issue is that when you run your scripts from init or from cron, they are not running within the GUI environment you want them to control. In principle, a Linux system can have multiple X environments running. When you are using one, there are environment variables that direct the executables you are using to the environment you are in.
There are two parts to the solution: your scripts have to know which environment they are acting on, and they have to have authorization to interact with that environment.
You almost certainly are using a DISPLAY value of ":0", so export DISPLAY=:0 at the beginning of your script will handle the first part of the problem. (It might be ":0.0", which is effectively equivalent).
Authorization is a bit more complex. X can be set up to do authorization in different ways, but the most common is to have a file .Xauthority in your home directory which contains a token that is checked by the X server. If you install a script in your own crontab, it will run under your own user id (you probabl shouldn't use sudo), so it will read the right .Xauthority file. If you run from the root crontab, or from an init script, it will run as the root user, so it will have access to everything but will still need to know where to take the token from. I think that adding export XAUTHORITY=/home/joe/.Xauthority to the script will work. (Assuming your user id is joe.)
I have 2 Amazon Linux EC2 instances that are running HAProxy. I want to monitor each instance from the other instance and if a instance becomes unavailable, the other instance will issue a API command to move the elastic IP to the active server.
I created a Bash script to do the monitoring every XX seconds. I need to set the script to run as a service so I created a service wrapper and placed in /etc/init.d based on a template that I found and registered as a service.
The problem is when I issue command #service hamonitor start, it says "Starting hamonitor...", but I never see the OK message and if I issue the stop command, it fails and if I issue the status command, it says it is not running. But, if I check the logs, it shows that the script is in fact running. I assume that I need a proper PID file and/or since the script runs in a infinite loop, it never completes so the OK does not get issued.
Service Wrapper:
#!/bin/sh
#
# /etc/init.d/hamonitor
# Subsystem file for "hamonitor" server
#
# chkconfig: 2345 95 05 (1)
# description: hamonitor server daemon
#
# processname: hamonitor
### BEGIN INIT INFO
# Provides:
# Required-Start:
# Required-Stop:
# Should-Start:
# Should-Stop:
# Default-Start:
# Default-Stop:
# Short-Description:
# Description:
### END INIT INFO
# source function library
. /etc/rc.d/init.d/functions
PROG=hamonitor
EXEC=/etc/haproxy/hamonitor
LOCKFILE=/var/lock/subsys/$prog
PIDFILE=/var/run/$prog.pid
RETVAL=0
start() {
echo -n $"Starting $PROG:"
echo
#daemon $EXEC &
/etc/haproxy/hamonitor &
RETVAL=$?
if [ $RETVAL -eq 0 ]; then
touch LOCKFILE
touch PIDFILE
echo "[ OK ]"
else
echo "[ FAIL: ${retval} ]"
fi
return $RETVAL
}
stop() {
echo -n $"Stopping $PROG:"
echo
killproc $PROG -TERM
RETVAL=$?
if [ $RETVAL -eq 0 ]; then
rm -f LOCKFILE
rm -f PIDFILE
echo "[ OK ]"
else
echo "[ FAIL: ${RETVAL} ]"
fi
return $RETVAL
}
case "$1" in
start)
start
;;
stop)
stop
;;
status)
status $PROG
RETVAL=$?
;;
restart)
stop
start
;;
*)
echo $"Usage: $0 {start|stop|status|restart}"
RETVAL=1
esac
exit $RETVAL
App:
#!/usr/bin/env bash
export EC2_HOME=/opt/aws/apitools/ec2
export JAVA_HOME=/usr/lib/jvm/jre
AWS_ACCESS_KEY="XXXXXXXXXXXXXXXXXXXXXXXXX"
AWS_SECRET_KEY="XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
VIP1="1.2.3.4"
VIP1_ALLOCATIONID="eipalloc-XXXXXXX"
THIS_NODE_EC2_ID="i-XXXXXXX"
THIS_NODE_PRIVATE_IPADDRESS1="10.60.0.11"
THIS_NODE_HEALTHCHECK_URL="http://10.60.0.10/haproxy?monitor"
OTHER_NODE_HEALTHCHECK_URL="http://10.60.49.50/haproxy?monitor"
CHECK_OTHER_INTERVAL=5
CHECK_OTHER_FAIL_COUNT=0
CHECK_OTHER_RUN_COUNT=0
AFTER_TAKEOVER_WAIT=30
function takeover_vips {
/opt/aws/bin/ec2-associate-address -aws-access-key ${AWS_ACCESS_KEY} -aws-secret-key ${AWS_SECRET_KEY} -a ${VIP1_ALLOCATIONID} -i ${THIS_NODE_EC2_ID} -private-ip-address ${THIS_NODE_PRIVATE_IPADDRESS1} -allow-reassociation > /dev/null
}
function does_this_node_have_ips {
is_active=$(/opt/aws/bin/ec2-describe-addresses -aws-access-key ${AWS_ACCESS_KEY} -aws-secret-key ${AWS_SECRET_KEY} | grep ${VIP1} | grep ${THIS_NODE_EC2_ID})
if [ "$is_active" = "" ]; then
echo "no"
else
echo "yes"
fi
}
function log_msg {
msg=$1
msg="$(date) -- ${msg}"
echo ${msg} >> /var/log/hamonitorlog
}
while [ . ]; do
healthcheck_response=$(curl -sL -w "%{http_code}" ${OTHER_NODE_HEALTHCHECK_URL} -o /dev/null)
if [ "$healthcheck_response" != "200" ]; then
CHECK_OTHER_FAIL_COUNT=$((CHECK_OTHER_FAIL_COUNT+1))
if [ "$CHECK_OTHER_FAIL_COUNT" -gt 2 ]; then
takeover_vips
CHECK_OTHER_FAIL_COUNT=0
sleep ${AFTER_TAKEOVER_WAIT}
fi
sleep ${CHECK_OTHER_INTERVAL}
done
Some Linux distribution have up-start and other init; I assume you have init. The chkconfig is being used to maintain symlinks. You should confirm the comment,
# chkconfig: 2345 95 05 (1)
is correct for your system.
As a guess, you need daemon to be invoked via a script. This may have been a script function in some init script library, like /etc/rc.d/init.d/functions. I would suggest that you use the daemon() function if it exists. Either,
daemon $EXEC & #option1
nohup /etc/haproxy/hamonitor < /dev/null > /dev/null 2>&1 & #option2
/etc/haproxy/hamonitor& #option3, 2 lines.
disown $! #...
This is related to SIGCHLD and process return status (see man wait for more). As well, you may need to detach hamonitor from the controlling terminal. You can use logger to send information to the system logs in this case; I guess the App script is the hamonitor code? Just change echo to logger.
If the hamonitor needs stdout, stdin, and/or stderr, you may need to redirect to some other file if it requires it. You might also consider running it via screen if this is the case.
Edit: The last option can be used to create a proper PIDFILE. For instance,
# !!! optional grabbing of lock here...
/etc/haproxy/hamonitor & # spawn in bg
HA_PID=$! # record spawn pid
echo $HA_PID > $PIDFILE # record the PID to a file for `stop`.
# !!! optional release of lock here...
disown $HA_PID # detach script from terminal.
Services should never use echo and the like; logger is the better option. This is probably not your issue unless hamonitor tries to read from something. Mainly the issue is that start() will wait for the hamonitor to finish if you don't disown, so the rc script's start will never finish.
Generically, you can look at /etc/rc.d/init.d/functions, provide a link to your file, or provide your distribution and version (or at least linux standard base conformance which seems to define how this should work in its different versions). The file can be different on each and every Linux. You can look at this file yourself if you understand scripting to see what environment variables, files, etc are expected and what functions you use in this file. For instance, killproc is most likely defined there.
I need to exit from the current session and I am using the following code:
read -p "Do you want to start a new session? [Y/N] " usr_session
if [ "$usr_session" == "y" ] || [ "$usr_session" == "Y" ]; then
echo -e "`date`\t\t Exiting...\n You will need to login back...\n" >> $LOG_FILE
echo -e "Exiting...\n You will need to login back...\n"
sleep 5
curr_usr=`whoami`
pkill -9 -u $curr_usr
elif [ "$usr_session" == "n" ] || [ "usr_session" == "N" ]; then
echo -e "You are still in the same session.\n"
else
echo "Invalid input"
fi
Is there a better approach to perform the same but in more graceful manner? I feel the killing the current user process might not be safe. Any suggestions are really appreciated.
Thanks in advance.
Consider using logout.
(If you don't care to immediately terminate all running processes, including background jobs)
You could send a "softer" signal first instead of -9, to give time for the processes to shut down gracefully, then only send -9 to the processes that won't exit.
logout will work only if you are able to run this script as part of the login script (such as bashrc), not if it's run in a subshell. Alternatively (with the same effect), run the script with exec script_name from the login shell.
Another idea is to kill -HUP $PPID, assuming that the script is always run directly as a subshell of the login shell. This will signal the parent shell (the login shell) to end the session.
you could write a wrapper script and then source in , for example
file : wrap1.sh
./myscript.sh
if [ $? != 0 ];then
exit
fi
replace kill with exit 1 and from main session :
source wrap1.sh
I am using an Amazon Linux AMI and doing some custom modifications(added an axis2server, etc) on it and saving it as a new AMI. Now what I want to do is when the AMI boots up, start up axis2server(ie.axis2server should automatically start when the instance boots up). For that I used a init script like below and ran the following command:
chkconfig --add axisservice
But when I launch a new instance from my image, the axis2server is not getting started.
I just only need to execute the script /home/ec2-user/axis2-1.6.1/bin/axis2server.sh at startup. Am I missing anything here?
#! /bin/sh
# Basic support for IRIX style chkconfig
###
# chkconfig: 235 98 55
# description: Manages the services you are controlling with the chkconfig command
###
case "$1" in
start)
echo -n "Starting axisservice"
touch ~/temp.txt
cd /home/ec2-user/axis2-1.6.1/bin
./axis2server.sh &
echo "."
;;
stop)
echo -n "Stopping axisservice"
echo "."
;;
*)
echo "Usage: /sbin/service axisservice {start|stop}"
exit 1
esac
exit 0
I went through https://help.ubuntu.com/community/CloudInit as well and it provides a mechanism called User-Data Scripts, where a user can execute a script when launching the script.
$ euca-run-instances --key mykey --user-data-file myscript.sh ami-axxxx
This is a command line option and what I want is something like when I launch the instance through the UI, the script should be started.Therefore, I think the above option can not be used in my case. Please correct me if I am wrong.
Thanks,
H.
I bet the environment is not set(up correctly). This means that I am guessing that your shell script tries to start another program and it's not to be found.
So at first, I'd adjust the start part of your script (current):
echo -n "Starting axisservice"
touch ~/temp.txt
cd /home/ec2-user/axis2-1.6.1/bin
./axis2server.sh &
echo "."
Edited:
echo -n "Starting axisservice"
touch ~/temp.txt
cd /home/ec2-user/axis2-1.6.1/bin
./axis2server.sh
RETVAL=$?
[ $RETVAL -eq 0 ] && echo Success
[ $RETVAL -ne 0 ] && echo Failure
echo "."
So what did I do?
removed & so script waits for your shell script (axis2server.sh) to complete
checked the return status ($?) of your shell script
Further debugging:
Add set -x to your scripts to enable tracing and log both stderr and stdout.
Questions:
Are you are aware that stop (in your service script) doesn't do anything?
touch ~/temp.txt is that supposed to create /root/temp.txt? (I'm guessing root runs this script.)
If none of my suggestions work, can you share axis2server.sh and paste stderr and stdout?