Undesirable change in 2.28 when there is no host output. about pdsh

[GoogleCodeExporter]

Thanks, this is a good find. It will be a severe bug for us as well because
running pdsh where there is output from only some nodes is a common idiom for
us.

Do you know if this issue only happens with rsh? (Have you tried mrsh/ssh/exec)

mark

Original comment by [email protected] on 11 Feb 2013 at 9:58

• Added labels: Priority-High
• Removed labels: Priority-Medium

[GoogleCodeExporter]

Ah, the -S option appears to be required to recreate this problem:

grondo@hype355:~/git/pdsh.git/src/pdsh$ pdsh -S -w hype[210-213] cd ..
hype210: hype212: hype211: hype213: grondo@hype355:~/git/pdsh.git/src/pdsh$ 
grondo@hype355:~/git/pdsh.git/src/pdsh$ pdsh -w hype[210-213] cd ..

and I cannot recreate the problem while using the 'exec' module (which is 
unfortunate
for writing a testcase, but another good clue here)

To collect return codes from remote processes the rsh and mrsh modules use a 
special
string embedded at the beginning of a line. If there is otherwise no output, 
then
when this string is removed it leaves an empty line sans newline, so we get the
unwanted behavior.

I'll try to find exactly when this broke.

Original comment by [email protected] on 11 Feb 2013 at 10:15

[GoogleCodeExporter]

So far, only tested with mrsh. Do not know if it happens with rsh or ssh.

Original comment by [email protected] on 11 Feb 2013 at 10:50

[GoogleCodeExporter]

Yes, git bisect verified your guess above that commit 
c37f0b533b93c0d6825df83dc3035ec54a1ce3e8 was the culprit.
Now comparing that change with preexisting code, I do see
that an if (strlen(buf) > 0) was dropped.

Thanks, this will be fixed in pdsh-2.29

Original comment by [email protected] on 11 Feb 2013 at 11:16

[GoogleCodeExporter]

This issue was closed by revision 60b39484c3c9.

Original comment by [email protected] on 11 Feb 2013 at 11:20

• Changed state: Fixed

[GoogleCodeExporter]

Joshua:

If you are willing can you verify the fix applied in revision 60b39484c3c9?

Original comment by [email protected] on 11 Feb 2013 at 11:26

[GoogleCodeExporter]

Just built and tested on el5 {i686|x86_64} and el6 {i686|x86_64}. Did not test 
suse11, but I can't imagine why it would be any different. The behavior is as 
expected.

How soon do you think this will be made into a release?

Original comment by [email protected] on 12 Feb 2013 at 2:20

[GoogleCodeExporter]

pdsh-2.29 should be out within the next day or so.

Original comment by [email protected] on 13 Feb 2013 at 12:18

[GoogleCodeExporter]

I've uploaded pdsh-2.29

http://pdsh.googlecode.com/files/pdsh-2.29.tar.bz2

Original comment by [email protected] on 13 Feb 2013 at 6:06