[SOLVED] [bash] consolidate multiple ssh commands into 1

schneidz · 04-04-2013, 01:23 PM

hi, i have a bash script that contains this:

Code:

...
f=`ssh hyper ls -1 /path/to/some/file | tail -n 1`
f=`ssh hyper grep $argv1 $f | cut -d \| -f 7 | grep hello-world | egrep -v "(l33t|h4x0rz)" | uniq`
fn=$(echo $f.gz | sed s/chun-li/chun-li\\/akuma/)
scp $USER@hyper:$fn .
...

seems kinda' inefficient to call ssh twice... i tired embedding them using `backticks`, "'quotes'", (parenthesis) but nothing seems to work. is this possible or should i just live with it ?

thanks,
schneidz

mina86 · 04-04-2013, 02:00 PM

Code:

f=`ssh hyper ls -1 /path/to/some/file | tail -n 1`

How does the effoct of the above line differ from:

Code:

f=/path/to/some/file

schneidz · 04-04-2013, 02:46 PM

Quote:

Originally Posted by mina86

Code:

f=`ssh hyper ls -1 /path/to/some/file | tail -n 1`

How does the effoct of the above line differ from:

Code:

f=/path/to/some/file

hi mina, becuz the last file in that directory will be date-stamped every friday (e.g.- log_2013-03-29.txt , last week the file would be log_2013-03-22.txt).

fyi, this is on an aix machine.

grail · 04-04-2013, 06:37 PM

Well my first impression is why do you use both `` and $()? I personally prefer the second but it does not seem to make it easy to follow if you change in and out of the two.

As for calling the ssh once, I believe the parenthesis option should work:

Code:

f=$(ssh hyper grep $argv1 $(ls -1 /path/to/some/file | tail -n 1) | cut -d \| -f 7 | grep hello-world | egrep -v "(l33t|h4x0rz)" | uniq)

It would appear that the several greps and cut seem a bit excessive. I know it is aix but I would have thought even its awk would do this a bit cleaner.

konsolebox · 04-05-2013, 12:09 AM

Quote:

Originally Posted by grail

Code:

$(ls -1 /path/to/some/file | tail -n 1)

I don't know why the OP has to but won't the command be not called remotely for that?

@schneidz

I think you could try sending a script or multiple commands to bash:

Code:

ssh hyper bash < script.sh
cat script.sh | ssh hyper bash
echo "grep \"$argv1\" \$(ls -1 /path/to/some/file | tail -n 1) | cut -d | -f 7 | grep hello-world | egrep -v \"(l33t|h4x0rz)\" | uniq" | ssh hyper bash
ssh hyper bash <<< "grep \"$argv1\" \$(ls -1 /path/to/some/file | tail -n 1) | cut -d | -f 7 | grep hello-world | egrep -v \"(l33t|h4x0rz)\" | uniq"

And place any of those statements in f=$(). Or if you could perhaps in fn=$() already. You should just note that everything sent to remote bash would be executed remotely as is.

Add: Sometimes you could also just use other shells. You just have to make your commands would work with them.

grail · 04-05-2013, 12:36 AM

Quote:

I don't know why the OP has to but won't the command be not called remotely for that?

I thought that was the point, that all commands would be executed remotely?

konsolebox · 04-05-2013, 03:31 AM

Quote:

Originally Posted by grail

I thought that was the point, that all commands would be executed remotely?

Yeah and what I meant was that the command won't be called remotely but on the shell that calls ssh.

mina86 · 04-05-2013, 07:34 AM

Quote:

Originally Posted by schneidz

hi mina, becuz the last file in that directory will be date-stamped every friday (e.g.- log_2013-03-29.txt , last week the file would be log_2013-03-22.txt).

Ah, so /path/to/some/file is really /path/to/some/directory.

You might try this:

Code:

fn=$(cat '
f=$(ls -1 /path/to/some/file | tail -n 1)
f=$(grep "$1" "$f" | cut -d \| -f 7 | grep hello-world | egrep -v "(l33t|h4x0rz)" | uniq)
echo $(echo "$f.gz" | sed s/chun-li/chun-li\\/akuma/)
' | ssh hyper sh -s "$argv1")

Not tested, not sure if -s is POSIX compliant.

schneidz · 04-05-2013, 08:39 AM

Quote:

Originally Posted by grail

Well my first impression is why do you use both `` and $()? I personally prefer the second but it does not seem to make it easy to follow if you change in and out of the two.

As for calling the ssh once, I believe the parenthesis option should work:

Code:

f=$(ssh hyper grep $argv1 $(ls -1 /path/to/some/file | tail -n 1) | cut -d \| -f 7 | grep hello-world | egrep -v "(l33t|h4x0rz)" | uniq)

It would appear that the several greps and cut seem a bit excessive. I know it is aix but I would have thought even its awk would do this a bit cleaner.

i think with this i got path not found errors becuz it was looking for /path/to/some/file on the local machine (which doesnt exists).

i think konsolebox's idea of piping into ssh looks promising.

just had another clever idea; maybe running find with -mtime -7 (each friday) and -exec cut ... grep ... uniq. maybe i can even cat the contents of the scraped filename and redirect it to a file on the local machine rather than calling scp independently.

grail · 04-05-2013, 09:44 AM

Sorry konsole ... my bad ... i get you now