I am experiencing an issue that sounds suspiciously like the hang
described above.  Unfortunately, changing the sshd configuration
(i.e., set AddressFamily to inet in sshd_config file) did not
alleviate the hang.

Specifically, I have a process that runs in my production environment
that periodically builds a tar file and then copies it to itself and
two other servers in the environment.  The upload is implemented using
the "put" method on the Capistrano::Actor class.  Roughly every other
time that this routine executes, the upload to itself and one other
server succeeds, but the upload to the third server hangs.  If you
look at the Capistrano output, you see the following:

(The file is being copied from xxx.xxx.xxx.xx4.)

  * uploading /u/ea_index_build/deploy/foo.tar.gz
 ** uploading data to xxx.xxx.xxx.xx2:/u/ea_index_build/deploy/
foo.tar.gz
 ** uploading data to xxx.xxx.xxx.xx3:/u/ea_index_build/deploy/
foo.tar.gz
 ** uploading data to xxx.xxx.xxx.xx4:/u/ea_index_build/deploy/
foo.tar.gz
  * done uploading data to xxx.xxx.xxx.xx3:/u/ea_index_build/deploy/
foo.tar.gz
  * done uploading data to xxx.xxx.xxx.xx4:/u/ea_index_build/deploy/
foo.tar.gz

Note that the upload to one of the other servers does not complete.
In this case, the server is xxx.xxx.xxx.xx2.  (It alternates between
xxx.xxx.xxx.xx2 and xxx.xxx.xxx.xx3.)  When you look at
xxx.xxx.xxx.xx2, you see that the file has been created and has a size
of 0 bytes.

Things that I have ruled out:

1. It is not server dependent.  The upload has failed for both
xxx.xxx.xxx.xx2 and xxx.xxx.xxx.xx3.  Which one fails tends to be
fairly random.  Also, it does not always fail.  Sometimes, the upload
to all of the servers works flawlessly.

2. It is not dependent upon the file being uploaded.  There are four
different files that can be uploaded by this process.  On differnet
occassions, each file has experienced the hang.

3. It does not appear to be related to the IPv6 issue.  I modified the
sshd_config on all three servers to only use "inet" as the
AddressFamily.  According to the doc, this is equivalent to specifying
"-4" on the command line for sshd.

Also, one last thing to note.  When the hang occurs, the server that
is running this process (xxx.xxx.xxx.xx4) shows a Ruby process taking
up ~100% of the CPU (using top to monitor).  The hang never returns.
When it does occur, I need to pkill the processes and restart things.

Well, that is my story.  I would be very interested to know if other
folks are running into this issue.  If you happen to have fixed or
worked around this issue, I would love to hear about it.

Thanks in advance for your help.

- Chuck

P.S., Jamis, thank you for creating Capistrano.  It rules!!!


--~--~---------~--~----~------------~-------~--~----~
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/capistrano
-~----------~----~----~----~------~----~------~--~---

Reply via email to