I am using a interactive command line program in a Linux terminal running the bash shell. I have a definite sequence of command that I input to the shell program. The program writes its output to standard output. One of these commands is a 'save' command, that writes the output of the previous command that was run, to a file to disk.
A typical cycle is:
$prog
$$cmdx
$$<some output>
$$save <filename>
$$cmdy
$$<again, some output>
$$save <filename>
$$q
$<back to bash shell>
$ is the bash prompt
$$ is the program's prompt
q is the quit command for prog
prog is such that it appends the output of the previous command to filename
How can I automate this process? I would like to write a shell script that can start this program, and cycle through the steps, feeding it the commands one by one and, and then quitting. I hope the save command works correctly.
If your command doesn't care how fast you give it input, and you don't really need to interact with it, then you can use a heredoc.
Example:
#!/bin/bash
prog <<EOD
cmdx
save filex
cmdy
save filey
q
EOD
If you need branching based on the output of the program, or if your program is at all sensitive to the timing of your commands, then Expect is what you want.
I recommend you use Expect. This tool is designed to automate interactive shell applications.
Where there's a need, there's a way! I think that it's a good bash lesson to see
how process management and ipc works. The best solution is, of course, Expect.
But the real reason is that pipes can be tricky and many commands are designed
to wait for data, meaning that the process will become a zombie for reasons that
bay be difficult to predict. But learning how and why reminds us of what is
going on under the hood.
When two processes engage in a conversation, the danger is that one or both will
try to read data that will never arrive. The rules of engagement have to be
crystal clear. Things like CRLF and character encoding can kill the party.
Luckily, two close partners like a bash script and its child process are
relatively easy to keep in line. The easiest thing to miss is that bash is
launching a child process for just about every thing it does. If you can make it
work with bash, you thoroughly know what you're doing.
The point is that we want to talk to another process. Here's a server:
# a really bad SMTP server
# a hint at courtesy to the client
shopt -s nocasematch
echo "220 $HOSTNAME SMTP [$$]"
while true
do
read
[[ "$REPLY" =~ ^helo\ [^\ ] ]] && break
[[ "$REPLY" =~ ^quit ]] && echo "Later" && exit
echo 503 5.5.1 Nice guys say hello.
done
NAME=`echo "$REPLY" | sed -r -e 's/^helo //i'`
echo 250 Hello there, $NAME
while read
do
[[ "$REPLY" =~ ^mail\ from: ]] && { echo 250 2.1.0 Good guess...; continue; }
[[ "$REPLY" =~ ^rcpt\ to: ]] && { echo 250 2.1.0 Keep trying...; continue; }
[[ "$REPLY" =~ ^quit ]] && { echo Later, $NAME; exit; }
echo 502 5.5.2 Please just QUIT
done
echo Pipe closed, exiting
Now, the script that hopefully does the magic.
# Talk to a subprocess using named pipes
rm -fr A B # don't use old pipes
mkfifo A B
# server will listen to A and send to B
./smtp.sh < A > B &
# If we write to A, the pipe will be closed.
# That doesn't happen when writing to a file handle.
exec 3>A
read < B
echo "$REPLY"
# send an email, so long as response codes look good
while read L
do
echo "> $L"
echo $L > A
read < B
echo $REPLY
[[ "$REPLY" =~ ^2 ]] || break
done <<EOF
HELO me
MAIL FROM: me
RCPT TO: you
DATA
Subject: Nothing
Message
.
EOF
# This is tricky, and the reason sane people use Expect. If we
# send QUIT and then wait on B (ie. cat B) we may have trouble.
# If the server exits, the "Later" response in the pipe might
# disappear, leaving the cat command (and us) waiting for data.
# So, let cat have our STDOUT and move on.
cat B &
# Now, we should wait for the cat process to get going before we
# send the QUIT command. If we don't, the server will exit, the
# pipe will empty and cat will miss its chance to show the
# server's final words.
echo -n > B # also, 'sleep 1' will probably work.
echo "> quit"
echo "quit" > A
# close the file handle
exec 3>&-
rm A B
Notice that we are not simply dumping the SMTP commands on the server. We check
each response code to make sure things are OK. In this case, things will not be
OK and the script will bail.
I use Expect to interact with the shell for switch and router backups. A bash script calls the expect script with the correct variables.
for i in <list of machines> ; do expect_script.sh $i ; exit
This will ssh to each box, run the backup commands, copy out the appropriate files, and then move on to the next box.
For simple use cases you may use a combination of subshell, echo & sleep:
# in Terminal.app
telnet localhost 25
helo localhost
ehlo localhost
quit
(sleep 5; echo "helo localhost"; sleep 5; echo "ehlo localhost"; sleep 5; echo quit ) |
telnet localhost 25
echo "cmdx\nsave\n...etc..." | prog
..?
Related
I want to execute a long running command in Bash, and both capture its exit status, and tee its output.
So I do this:
command | tee out.txt
ST=$?
The problem is that the variable ST captures the exit status of tee and not of command. How can I solve this?
Note that command is long running and redirecting the output to a file to view it later is not a good solution for me.
There is an internal Bash variable called $PIPESTATUS; it’s an array that holds the exit status of each command in your last foreground pipeline of commands.
<command> | tee out.txt ; test ${PIPESTATUS[0]} -eq 0
Or another alternative which also works with other shells (like zsh) would be to enable pipefail:
set -o pipefail
...
The first option does not work with zsh due to a little bit different syntax.
Dumb solution: Connecting them through a named pipe (mkfifo). Then the command can be run second.
mkfifo pipe
tee out.txt < pipe &
command > pipe
echo $?
using bash's set -o pipefail is helpful
pipefail: the return value of a pipeline is the status of
the last command to exit with a non-zero status,
or zero if no command exited with a non-zero status
There's an array that gives you the exit status of each command in a pipe.
$ cat x| sed 's///'
cat: x: No such file or directory
$ echo $?
0
$ cat x| sed 's///'
cat: x: No such file or directory
$ echo ${PIPESTATUS[*]}
1 0
$ touch x
$ cat x| sed 's'
sed: 1: "s": substitute pattern can not be delimited by newline or backslash
$ echo ${PIPESTATUS[*]}
0 1
This solution works without using bash specific features or temporary files. Bonus: in the end the exit status is actually an exit status and not some string in a file.
Situation:
someprog | filter
you want the exit status from someprog and the output from filter.
Here is my solution:
((((someprog; echo $? >&3) | filter >&4) 3>&1) | (read xs; exit $xs)) 4>&1
echo $?
See my answer for the same question on unix.stackexchange.com for a detailed explanation and an alternative without subshells and some caveats.
By combining PIPESTATUS[0] and the result of executing the exit command in a subshell, you can directly access the return value of your initial command:
command | tee ; ( exit ${PIPESTATUS[0]} )
Here's an example:
# the "false" shell built-in command returns 1
false | tee ; ( exit ${PIPESTATUS[0]} )
echo "return value: $?"
will give you:
return value: 1
So I wanted to contribute an answer like lesmana's, but I think mine is perhaps a little simpler and slightly more advantageous pure-Bourne-shell solution:
# You want to pipe command1 through command2:
exec 4>&1
exitstatus=`{ { command1; printf $? 1>&3; } | command2 1>&4; } 3>&1`
# $exitstatus now has command1's exit status.
I think this is best explained from the inside out - command1 will execute and print its regular output on stdout (file descriptor 1), then once it's done, printf will execute and print icommand1's exit code on its stdout, but that stdout is redirected to file descriptor 3.
While command1 is running, its stdout is being piped to command2 (printf's output never makes it to command2 because we send it to file descriptor 3 instead of 1, which is what the pipe reads). Then we redirect command2's output to file descriptor 4, so that it also stays out of file descriptor 1 - because we want file descriptor 1 free for a little bit later, because we will bring the printf output on file descriptor 3 back down into file descriptor 1 - because that's what the command substitution (the backticks), will capture and that's what will get placed into the variable.
The final bit of magic is that first exec 4>&1 we did as a separate command - it opens file descriptor 4 as a copy of the external shell's stdout. Command substitution will capture whatever is written on standard out from the perspective of the commands inside it - but since command2's output is going to file descriptor 4 as far as the command substitution is concerned, the command substitution doesn't capture it - however once it gets "out" of the command substitution it is effectively still going to the script's overall file descriptor 1.
(The exec 4>&1 has to be a separate command because many common shells don't like it when you try to write to a file descriptor inside a command substitution, that is opened in the "external" command that is using the substitution. So this is the simplest portable way to do it.)
You can look at it in a less technical and more playful way, as if the outputs of the commands are leapfrogging each other: command1 pipes to command2, then the printf's output jumps over command 2 so that command2 doesn't catch it, and then command 2's output jumps over and out of the command substitution just as printf lands just in time to get captured by the substitution so that it ends up in the variable, and command2's output goes on its merry way being written to the standard output, just as in a normal pipe.
Also, as I understand it, $? will still contain the return code of the second command in the pipe, because variable assignments, command substitutions, and compound commands are all effectively transparent to the return code of the command inside them, so the return status of command2 should get propagated out - this, and not having to define an additional function, is why I think this might be a somewhat better solution than the one proposed by lesmana.
Per the caveats lesmana mentions, it's possible that command1 will at some point end up using file descriptors 3 or 4, so to be more robust, you would do:
exec 4>&1
exitstatus=`{ { command1 3>&-; printf $? 1>&3; } 4>&- | command2 1>&4; } 3>&1`
exec 4>&-
Note that I use compound commands in my example, but subshells (using ( ) instead of { } will also work, though may perhaps be less efficient.)
Commands inherit file descriptors from the process that launches them, so the entire second line will inherit file descriptor four, and the compound command followed by 3>&1 will inherit the file descriptor three. So the 4>&- makes sure that the inner compound command will not inherit file descriptor four, and the 3>&- will not inherit file descriptor three, so command1 gets a 'cleaner', more standard environment. You could also move the inner 4>&- next to the 3>&-, but I figure why not just limit its scope as much as possible.
I'm not sure how often things use file descriptor three and four directly - I think most of the time programs use syscalls that return not-used-at-the-moment file descriptors, but sometimes code writes to file descriptor 3 directly, I guess (I could imagine a program checking a file descriptor to see if it's open, and using it if it is, or behaving differently accordingly if it's not). So the latter is probably best to keep in mind and use for general-purpose cases.
(command | tee out.txt; exit ${PIPESTATUS[0]})
Unlike #cODAR's answer this returns the original exit code of the first command and not only 0 for success and 127 for failure. But as #Chaoran pointed out you can just call ${PIPESTATUS[0]}. It is important however that all is put into brackets.
In Ubuntu and Debian, you can apt-get install moreutils. This contains a utility called mispipe that returns the exit status of the first command in the pipe.
Outside of bash, you can do:
bash -o pipefail -c "command1 | tee output"
This is useful for example in ninja scripts where the shell is expected to be /bin/sh.
The simplest way to do this in plain bash is to use process substitution instead of a pipeline. There are several differences, but they probably don't matter very much for your use case:
When running a pipeline, bash waits until all processes complete.
Sending Ctrl-C to bash makes it kill all the processes of a pipeline, not just the main one.
The pipefail option and the PIPESTATUS variable are irrelevant to process substitution.
Possibly more
With process substitution, bash just starts the process and forgets about it, it's not even visible in jobs.
Mentioned differences aside, consumer < <(producer) and producer | consumer are essentially equivalent.
If you want to flip which one is the "main" process, you just flip the commands and the direction of the substitution to producer > >(consumer). In your case:
command > >(tee out.txt)
Example:
$ { echo "hello world"; false; } > >(tee out.txt)
hello world
$ echo $?
1
$ cat out.txt
hello world
$ echo "hello world" > >(tee out.txt)
hello world
$ echo $?
0
$ cat out.txt
hello world
As I said, there are differences from the pipe expression. The process may never stop running, unless it is sensitive to the pipe closing. In particular, it may keep writing things to your stdout, which may be confusing.
PIPESTATUS[#] must be copied to an array immediately after the pipe command returns.
Any reads of PIPESTATUS[#] will erase the contents.
Copy it to another array if you plan on checking the status of all pipe commands.
"$?" is the same value as the last element of "${PIPESTATUS[#]}",
and reading it seems to destroy "${PIPESTATUS[#]}", but I haven't absolutely verified this.
declare -a PSA
cmd1 | cmd2 | cmd3
PSA=( "${PIPESTATUS[#]}" )
This will not work if the pipe is in a sub-shell. For a solution to that problem,
see bash pipestatus in backticked command?
Base on #brian-s-wilson 's answer; this bash helper function:
pipestatus() {
local S=("${PIPESTATUS[#]}")
if test -n "$*"
then test "$*" = "${S[*]}"
else ! [[ "${S[#]}" =~ [^0\ ] ]]
fi
}
used thus:
1: get_bad_things must succeed, but it should produce no output; but we want to see output that it does produce
get_bad_things | grep '^'
pipeinfo 0 1 || return
2: all pipeline must succeed
thing | something -q | thingy
pipeinfo || return
Pure shell solution:
% rm -f error.flag; echo hello world \
| (cat || echo "First command failed: $?" >> error.flag) \
| (cat || echo "Second command failed: $?" >> error.flag) \
| (cat || echo "Third command failed: $?" >> error.flag) \
; test -s error.flag && (echo Some command failed: ; cat error.flag)
hello world
And now with the second cat replaced by false:
% rm -f error.flag; echo hello world \
| (cat || echo "First command failed: $?" >> error.flag) \
| (false || echo "Second command failed: $?" >> error.flag) \
| (cat || echo "Third command failed: $?" >> error.flag) \
; test -s error.flag && (echo Some command failed: ; cat error.flag)
Some command failed:
Second command failed: 1
First command failed: 141
Please note the first cat fails as well, because it's stdout gets closed on it. The order of the failed commands in the log is correct in this example, but don't rely on it.
This method allows for capturing stdout and stderr for the individual commands so you can then dump that as well into a log file if an error occurs, or just delete it if no error (like the output of dd).
It may sometimes be simpler and clearer to use an external command, rather than digging into the details of bash. pipeline, from the minimal process scripting language execline, exits with the return code of the second command*, just like a sh pipeline does, but unlike sh, it allows reversing the direction of the pipe, so that we can capture the return code of the producer process (the below is all on the sh command line, but with execline installed):
$ # using the full execline grammar with the execlineb parser:
$ execlineb -c 'pipeline { echo "hello world" } tee out.txt'
hello world
$ cat out.txt
hello world
$ # for these simple examples, one can forego the parser and just use "" as a separator
$ # traditional order
$ pipeline echo "hello world" "" tee out.txt
hello world
$ # "write" order (second command writes rather than reads)
$ pipeline -w tee out.txt "" echo "hello world"
hello world
$ # pipeline execs into the second command, so that's the RC we get
$ pipeline -w tee out.txt "" false; echo $?
1
$ pipeline -w tee out.txt "" true; echo $?
0
$ # output and exit status
$ pipeline -w tee out.txt "" sh -c "echo 'hello world'; exit 42"; echo "RC: $?"
hello world
RC: 42
$ cat out.txt
hello world
Using pipeline has the same differences to native bash pipelines as the bash process substitution used in answer #43972501.
* Actually pipeline doesn't exit at all unless there is an error. It executes into the second command, so it's the second command that does the returning.
Why not use stderr? Like so:
(
# Our long-running process that exits abnormally
( for i in {1..100} ; do echo ploop ; sleep 0.5 ; done ; exit 5 )
echo $? 1>&2 # We pass the exit status of our long-running process to stderr (fd 2).
) | tee ploop.out
So ploop.out receives the stdout. stderr receives the exit status of the long running process. This has the benefit of being completely POSIX-compatible.
(Well, with the exception of the range expression in the example long-running process, but that's not really relevant.)
Here's what this looks like:
...
ploop
ploop
ploop
ploop
ploop
ploop
ploop
ploop
ploop
ploop
5
Note that the return code 5 does not get output to the file ploop.out.
I have the following shell script. The purpose is to loop thru each line of the target file (whose path is the input parameter to the script) and do work against each line. Now, it seems only work with the very first line in the target file and stops after that line got processed. Is there anything wrong with my script?
#!/bin/bash
# SCRIPT: do.sh
# PURPOSE: loop thru the targets
FILENAME=$1
count=0
echo "proceed with $FILENAME"
while read LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done < $FILENAME
echo "\ntotal $count targets"
In do_work.sh, I run a couple of ssh commands.
The problem is that do_work.sh runs ssh commands and by default ssh reads from stdin which is your input file. As a result, you only see the first line processed, because the command consumes the rest of the file and your while loop terminates.
This happens not just for ssh, but for any command that reads stdin, including mplayer, ffmpeg, HandBrakeCLI, httpie, brew install, and more.
To prevent this, pass the -n option to your ssh command to make it read from /dev/null instead of stdin. Other commands have similar flags, or you can universally use < /dev/null.
A very simple and robust workaround is to change the file descriptor from which the read command receives input.
This is accomplished by two modifications: the -u argument to read, and the redirection operator for < $FILENAME.
In BASH, the default file descriptor values (i.e. values for -u in read) are:
0 = stdin
1 = stdout
2 = stderr
So just choose some other unused file descriptor, like 9 just for fun.
Thus, the following would be the workaround:
while read -u 9 LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done 9< $FILENAME
Notice the two modifications:
read becomes read -u 9
< $FILENAME becomes 9< $FILENAME
As a best practice, I do this for all while loops I write in BASH.
If you have nested loops using read, use a different file descriptor for each one (9,8,7,...).
More generally, a workaround which isn't specific to ssh is to redirect standard input for any command which might otherwise consume the while loop's input.
while read -r line; do
((count++))
echo "$count $line"
sh ./do_work.sh "$line" </dev/null
done < "$filename"
The addition of </dev/null is the crucial point here, though the corrected quoting is also somewhat important for robustness; see also When to wrap quotes around a shell variable?. You will want to use read -r unless you specifically require the slightly odd legacy behavior you get for backslashes in the input without -r. Finally, avoid upper case for your private variables.
Another workaround of sorts which is somewhat specific to ssh is to make sure any ssh command has its standard input tied up, e.g. by changing
ssh otherhost some commands here
to instead read the commands from a here document, which conveniently (for this particular scenario) ties up the standard input of ssh for the commands:
ssh otherhost <<'____HERE'
some commands here
____HERE
ssh -n option prevents checking the exit status of ssh when using HEREdoc while piping output to another program.
So use of /dev/null as stdin is preferred.
#!/bin/bash
while read ONELINE ; do
ssh ubuntu#host_xyz </dev/null <<EOF 2>&1 | filter_pgm
echo "Hi, $ONELINE. You come here often?"
process_response_pgm
EOF
if [ ${PIPESTATUS[0]} -ne 0 ] ; then
echo "aborting loop"
exit ${PIPESTATUS[0]}
fi
done << input_list.txt
This was happening to me because I had set -e and a grep in a loop was returning with no output (which gives a non-zero error code).
I have the following shell script. The purpose is to loop thru each line of the target file (whose path is the input parameter to the script) and do work against each line. Now, it seems only work with the very first line in the target file and stops after that line got processed. Is there anything wrong with my script?
#!/bin/bash
# SCRIPT: do.sh
# PURPOSE: loop thru the targets
FILENAME=$1
count=0
echo "proceed with $FILENAME"
while read LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done < $FILENAME
echo "\ntotal $count targets"
In do_work.sh, I run a couple of ssh commands.
The problem is that do_work.sh runs ssh commands and by default ssh reads from stdin which is your input file. As a result, you only see the first line processed, because the command consumes the rest of the file and your while loop terminates.
This happens not just for ssh, but for any command that reads stdin, including mplayer, ffmpeg, HandBrakeCLI, httpie, brew install, and more.
To prevent this, pass the -n option to your ssh command to make it read from /dev/null instead of stdin. Other commands have similar flags, or you can universally use < /dev/null.
A very simple and robust workaround is to change the file descriptor from which the read command receives input.
This is accomplished by two modifications: the -u argument to read, and the redirection operator for < $FILENAME.
In BASH, the default file descriptor values (i.e. values for -u in read) are:
0 = stdin
1 = stdout
2 = stderr
So just choose some other unused file descriptor, like 9 just for fun.
Thus, the following would be the workaround:
while read -u 9 LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done 9< $FILENAME
Notice the two modifications:
read becomes read -u 9
< $FILENAME becomes 9< $FILENAME
As a best practice, I do this for all while loops I write in BASH.
If you have nested loops using read, use a different file descriptor for each one (9,8,7,...).
More generally, a workaround which isn't specific to ssh is to redirect standard input for any command which might otherwise consume the while loop's input.
while read -r line; do
((count++))
echo "$count $line"
sh ./do_work.sh "$line" </dev/null
done < "$filename"
The addition of </dev/null is the crucial point here, though the corrected quoting is also somewhat important for robustness; see also When to wrap quotes around a shell variable?. You will want to use read -r unless you specifically require the slightly odd legacy behavior you get for backslashes in the input without -r. Finally, avoid upper case for your private variables.
Another workaround of sorts which is somewhat specific to ssh is to make sure any ssh command has its standard input tied up, e.g. by changing
ssh otherhost some commands here
to instead read the commands from a here document, which conveniently (for this particular scenario) ties up the standard input of ssh for the commands:
ssh otherhost <<'____HERE'
some commands here
____HERE
ssh -n option prevents checking the exit status of ssh when using HEREdoc while piping output to another program.
So use of /dev/null as stdin is preferred.
#!/bin/bash
while read ONELINE ; do
ssh ubuntu#host_xyz </dev/null <<EOF 2>&1 | filter_pgm
echo "Hi, $ONELINE. You come here often?"
process_response_pgm
EOF
if [ ${PIPESTATUS[0]} -ne 0 ] ; then
echo "aborting loop"
exit ${PIPESTATUS[0]}
fi
done << input_list.txt
This was happening to me because I had set -e and a grep in a loop was returning with no output (which gives a non-zero error code).
I have the following shell script. The purpose is to loop thru each line of the target file (whose path is the input parameter to the script) and do work against each line. Now, it seems only work with the very first line in the target file and stops after that line got processed. Is there anything wrong with my script?
#!/bin/bash
# SCRIPT: do.sh
# PURPOSE: loop thru the targets
FILENAME=$1
count=0
echo "proceed with $FILENAME"
while read LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done < $FILENAME
echo "\ntotal $count targets"
In do_work.sh, I run a couple of ssh commands.
The problem is that do_work.sh runs ssh commands and by default ssh reads from stdin which is your input file. As a result, you only see the first line processed, because the command consumes the rest of the file and your while loop terminates.
This happens not just for ssh, but for any command that reads stdin, including mplayer, ffmpeg, HandBrakeCLI, httpie, brew install, and more.
To prevent this, pass the -n option to your ssh command to make it read from /dev/null instead of stdin. Other commands have similar flags, or you can universally use < /dev/null.
A very simple and robust workaround is to change the file descriptor from which the read command receives input.
This is accomplished by two modifications: the -u argument to read, and the redirection operator for < $FILENAME.
In BASH, the default file descriptor values (i.e. values for -u in read) are:
0 = stdin
1 = stdout
2 = stderr
So just choose some other unused file descriptor, like 9 just for fun.
Thus, the following would be the workaround:
while read -u 9 LINE; do
let count++
echo "$count $LINE"
sh ./do_work.sh $LINE
done 9< $FILENAME
Notice the two modifications:
read becomes read -u 9
< $FILENAME becomes 9< $FILENAME
As a best practice, I do this for all while loops I write in BASH.
If you have nested loops using read, use a different file descriptor for each one (9,8,7,...).
More generally, a workaround which isn't specific to ssh is to redirect standard input for any command which might otherwise consume the while loop's input.
while read -r line; do
((count++))
echo "$count $line"
sh ./do_work.sh "$line" </dev/null
done < "$filename"
The addition of </dev/null is the crucial point here, though the corrected quoting is also somewhat important for robustness; see also When to wrap quotes around a shell variable?. You will want to use read -r unless you specifically require the slightly odd legacy behavior you get for backslashes in the input without -r. Finally, avoid upper case for your private variables.
Another workaround of sorts which is somewhat specific to ssh is to make sure any ssh command has its standard input tied up, e.g. by changing
ssh otherhost some commands here
to instead read the commands from a here document, which conveniently (for this particular scenario) ties up the standard input of ssh for the commands:
ssh otherhost <<'____HERE'
some commands here
____HERE
ssh -n option prevents checking the exit status of ssh when using HEREdoc while piping output to another program.
So use of /dev/null as stdin is preferred.
#!/bin/bash
while read ONELINE ; do
ssh ubuntu#host_xyz </dev/null <<EOF 2>&1 | filter_pgm
echo "Hi, $ONELINE. You come here often?"
process_response_pgm
EOF
if [ ${PIPESTATUS[0]} -ne 0 ] ; then
echo "aborting loop"
exit ${PIPESTATUS[0]}
fi
done << input_list.txt
This was happening to me because I had set -e and a grep in a loop was returning with no output (which gives a non-zero error code).
Did some search online, found simple 'tutorials' to use named pipes. However when I do anything with background jobs I seem to lose a lot of data.
[[Edit: found a much simpler solution, see reply to post. So the question I put forward is now academic - in case one might want a job server]]
Using Ubuntu 10.04 with Linux 2.6.32-25-generic #45-Ubuntu SMP Sat Oct 16 19:52:42 UTC 2010 x86_64 GNU/Linux
GNU bash, version 4.1.5(1)-release (x86_64-pc-linux-gnu).
My bash function is:
function jqs
{
pipe=/tmp/__job_control_manager__
trap "rm -f $pipe; exit" EXIT SIGKILL
if [[ ! -p "$pipe" ]]; then
mkfifo "$pipe"
fi
while true
do
if read txt <"$pipe"
then
echo "$(date +'%Y'): new text is [[$txt]]"
if [[ "$txt" == 'quit' ]]
then
break
fi
fi
done
}
I run this in the background:
> jqs&
[1] 5336
And now I feed it:
for i in 1 2 3 4 5 6 7 8
do
(echo aaa$i > /tmp/__job_control_manager__ && echo success$i &)
done
The output is inconsistent.
I frequently don't get all success echoes.
I get at most as many new text echos as success echoes, sometimes less.
If I remove the '&' from the 'feed', it seems to work, but I am blocked until the output is read. Hence me wanting to let sub-processes get blocked, but not the main process.
The aim being to write a simple job control script so I can run say 10 jobs in parallel at most and queue the rest for later processing, but reliably know that they do run.
Full job manager below:
function jq_manage
{
export __gn__="$1"
pipe=/tmp/__job_control_manager_"$__gn__"__
trap "rm -f $pipe" EXIT
trap "break" SIGKILL
if [[ ! -p "$pipe" ]]; then
mkfifo "$pipe"
fi
while true
do
date
jobs
if (($(jobs | egrep "Running.*echo '%#_Group_#%_$__gn__'" | wc -l) < $__jN__))
then
echo "Waiting for new job"
if read new_job <"$pipe"
then
echo "new job is [[$new_job]]"
if [[ "$new_job" == 'quit' ]]
then
break
fi
echo "In group $__gn__, starting job $new_job"
eval "(echo '%#_Group_#%_$__gn__' > /dev/null; $new_job) &"
fi
else
sleep 3
fi
done
}
function jq
{
# __gn__ = first parameter to this function, the job group name (the pool within which to allocate __jN__ jobs)
# __jN__ = second parameter to this function, the maximum of job numbers to run concurrently
export __gn__="$1"
shift
export __jN__="$1"
shift
export __jq__=$(jobs | egrep "Running.*echo '%#_GroupQueue_#%_$__gn__'" | wc -l)
if (($__jq__ '<' 1))
then
eval "(echo '%#_GroupQueue_#%_$__gn__' > /dev/null; jq_manage $__gn__) &"
fi
pipe=/tmp/__job_control_manager_"$__gn__"__
echo $# >$pipe
}
Calling
jq <name> <max processes> <command>
jq abc 2 sleep 20
will start one process.
That part works fine. Start a second one, fine.
One by one by hand seem to work fine.
But starting 10 in a loop seems to lose the system, as in the simpler example above.
Any hints as to what I can do to solve this apparent loss of IPC data would be greatly appreciated.
Regards,
Alain.
Your problem is if statement below:
while true
do
if read txt <"$pipe"
....
done
What is happening is that your job queue server is opening and closing the pipe each time around the loop. This means that some of the clients are getting a "broken pipe" error when they try to write to the pipe - that is, the reader of the pipe goes away after the writer opens it.
To fix this, change your loop in the server open the pipe once for the entire loop:
while true
do
if read txt
....
done < "$pipe"
Done this way, the pipe is opened once and kept open.
You will need to be careful of what you run inside the loop, as all processing inside the loop will have stdin attached to the named pipe. You will want to make sure you redirect stdin of all your processes inside the loop from somewhere else, otherwise they may consume the data from the pipe.
Edit: With the problem now being that you are getting EOF on your reads when the last client closes the pipe, you can use jilles method of duping the file descriptors, or you can just make sure you are a client too and keep the write side of the pipe open:
while true
do
if read txt
....
done < "$pipe" 3> "$pipe"
This will hold the write side of the pipe open on fd 3. The same caveat applies with this file descriptor as with stdin. You will need to close it so any child processes dont inherit it. It probably matters less than with stdin, but it would be cleaner.
As said in other answers you need to keep the fifo open at all times to avoid losing data.
However, once all writers have left after the fifo has been open (so there was a writer), reads return immediately (and poll() returns POLLHUP). The only way to clear this state is to reopen the fifo.
POSIX does not provide a solution to this but at least Linux and FreeBSD do: if reads start failing, open the fifo again while keeping the original descriptor open. This works because in Linux and FreeBSD the "hangup" state is local to a particular open file description, while in POSIX it is global to the fifo.
This can be done in a shell script like this:
while :; do
exec 3<tmp/testfifo
exec 4<&-
while read x; do
echo "input: $x"
done <&3
exec 4<&3
exec 3<&-
done
Just for those that might be interested, [[re-edited]] following comments by camh and jilles, here are two new versions of the test server script.
Both versions now works exactly as hoped.
camh's version for pipe management:
function jqs # Job queue manager
{
pipe=/tmp/__job_control_manager__
trap "rm -f $pipe; exit" EXIT TERM
if [[ ! -p "$pipe" ]]; then
mkfifo "$pipe"
fi
while true
do
if read -u 3 txt
then
echo "$(date +'%Y'): new text is [[$txt]]"
if [[ "$txt" == 'quit' ]]
then
break
else
sleep 1
# process $txt - remember that if this is to be a spawned job, we should close fd 3 and 4 beforehand
fi
fi
done 3< "$pipe" 4> "$pipe" # 4 is just to keep the pipe opened so any real client does not end up causing read to return EOF
}
jille's version for pipe management:
function jqs # Job queue manager
{
pipe=/tmp/__job_control_manager__
trap "rm -f $pipe; exit" EXIT TERM
if [[ ! -p "$pipe" ]]; then
mkfifo "$pipe"
fi
exec 3< "$pipe"
exec 4<&-
while true
do
if read -u 3 txt
then
echo "$(date +'%Y'): new text is [[$txt]]"
if [[ "$txt" == 'quit' ]]
then
break
else
sleep 1
# process $txt - remember that if this is to be a spawned job, we should close fd 3 and 4 beforehand
fi
else
# Close the pipe and reconnect it so that the next read does not end up returning EOF
exec 4<&3
exec 3<&-
exec 3< "$pipe"
exec 4<&-
fi
done
}
Thanks to all for your help.
Like camh & Dennis Williamson say don't break the pipe.
Now I have smaller examples, direct on the command line:
Server:
(
for i in {0,1,2,3,4}{0,1,2,3,4,5,6,7,8,9};
do
if read s;
then echo ">>$i--$s//";
else
echo "<<$i";
fi;
done < tst-fifo
)&
Client:
(
for i in {%a,#b}{1,2}{0,1};
do
echo "Test-$i" > tst-fifo;
done
)&
Can replace the key line with:
(echo "Test-$i" > tst-fifo&);
All client data sent to the pipe gets read, though with option two of the client one may need to start the server a couple of times before all data is read.
But although the read waits for data in the pipe to start with, once data has been pushed, it reads the empty string forever.
Any way to stop this?
Thanks for any insights again.
On the one hand the problem is worse than I thought:
Now there seems to be a case in my more complex example (jq_manage) where the same data is being read over and over again from the pipe (even though no new data is being written to it).
On the other hand, I found a simple solution (edited following Dennis' comment):
function jqn # compute the number of jobs running in that group
{
__jqty__=$(jobs | egrep "Running.*echo '%#_Group_#%_$__groupn__'" | wc -l)
}
function jq
{
__groupn__="$1"; shift # job group name (the pool within which to allocate $__jmax__ jobs)
__jmax__="$1"; shift # maximum of job numbers to run concurrently
jqn
while (($__jqty__ '>=' $__jmax__))
do
sleep 1
jqn
done
eval "(echo '%#_Group_#%_$__groupn__' > /dev/null; $#) &"
}
Works like a charm.
No socket or pipe involved.
Simple.
run say 10 jobs in parallel at most and queue the rest for later processing, but reliably know that they do run
You can do this with GNU Parallel. You will not need a this scripting.
http://www.gnu.org/software/parallel/man.html#options
You can set max-procs "Number of jobslots. Run up to N jobs in parallel." There is an option to set the number of CPU cores you want to use. You can save the list of executed jobs to a log file, but that is a beta feature.