From: Robert Klemme <shortcutter@...>
Date: 2011-11-17T00:21:27+09:00
Subject: Re: SSH hangs for 30 sec on AIX

On Wed, Nov 16, 2011 at 1:41 PM, Claus Folke Brobak <cfb@jndata.dk> wrote:
> Environment:
>
> - AIX 6.1
>
> - Java:
>  java version "1.6.0"
>  Java(TM) SE Runtime Environment (build pap3260sr9fp1-20110208_03(SR9
> FP1))
>  IBM J9 VM (build 2.4, JRE 1.6.0 IBM J9 2.4 AIX ppc-32
> jvmap3260sr9-20110203_74623 (JIT enabled, AOT enabled)
>  J9VM - 20110203_074623
>  JIT  - r9_20101028_17488ifx3
>  GC   - 20101027_AA)
>  JCL  - 20110203_01
>
> - jruby 1.6.5 (ruby-1.8.7-p330) (2011-10-25 9dcd388) (IBM J9 VM 1.6.0)
> [AIX-ppc-java]
>
> - Gems:
>  - bouncy-castle-java (1.5.0146.1)
>  - jruby-openssl (0.7.4)
>  - net-ssh (2.2.1)
>
> Code:
>    require 'rubygems'
>    require 'net/ssh'
>    host = 'xxxxxxx.xxxxxx.xxx'
>    username = 'xxxx'
>    Net::SSH.start(host, username, :verbose => :debug) do |ssh|
>      puts ssh.exec!("hostname")
>      puts ssh.exec!("ls")
>    end
>
> When running this program it hangs for 30 seconds at four points, as I
> have shown below. Note that I have omitted a number of lines.
>
>    -
>    -
>    D, [2011-11-16T11:53:11.477000 #1213352018] DEBUG -- tcpsocket[7d4]:
> received packet nr 6 type 52 len 12
>    D, [2011-11-16T11:53:11.478000 #1213352018] DEBUG --
> net.ssh.authentication.methods.publickey[7dc]: publickey succeeded
> (2d:32:27:ed:60:c8:28:28:77:50:7d:40:d7:83:ef:4b)
>    D, [2011-11-16T11:53:11.482000 #1213352018] DEBUG -- tcpsocket[7d4]:
> queueing packet nr 7 type 90 len 44
>    <<< Hangs for 30 seconds, then continues >>>
>    -
>    -
>    D, [2011-11-16T11:53:41.573000 #1213352018] DEBUG -- tcpsocket[7d4]:
> received packet nr 8 type 91 len 28
>    I, [2011-11-16T11:53:41.577000 #1213352018]  INFO --
> net.ssh.connection.session[7de]: channel_open_confirmation: 0 0 0 32768
>    I, [2011-11-16T11:53:41.579000 #1213352018]  INFO --
> net.ssh.connection.channel[7e0]: sending channel request "exec"
>    D, [2011-11-16T11:53:41.586000 #1213352018] DEBUG -- tcpsocket[7d4]:
> queueing packet nr 9 type 98 len 44
>    <<< Hangs for 30 seconds, then continues >>>
>    -
>    -
>    D, [2011-11-16T11:54:11.556000 #1213352018] DEBUG -- tcpsocket[7d4]:
> received packet nr 15 type 97 len 12
>    I, [2011-11-16T11:54:11.558000 #1213352018]  INFO --
> net.ssh.connection.session[7de]: channel_close: 0
>    D, [2011-11-16T11:54:11.560000 #1213352018] DEBUG -- tcpsocket[7d4]:
> queueing packet nr 11 type 97 len 28
>    <<< Output from hostname command >>>
>    D, [2011-11-16T11:54:11.563000 #1213352018] DEBUG -- tcpsocket[7d4]:
> queueing packet nr 12 type 90 len 44
>    <<< Hangs for 30 seconds, then continues >>>
>    -
>    -
>    D, [2011-11-16T11:54:41.510000 #1213352018] DEBUG -- tcpsocket[7d4]:
> received packet nr 17 type 91 len 28
>    I, [2011-11-16T11:54:41.514000 #1213352018]  INFO --
> net.ssh.connection.session[7de]: channel_open_confirmation: 1 1 0 32768
>    I, [2011-11-16T11:54:41.515000 #1213352018]  INFO --
> net.ssh.connection.channel[7e2]: sending channel request "exec"
>    D, [2011-11-16T11:54:41.518000 #1213352018] DEBUG -- tcpsocket[7d4]:
> queueing packet nr 13 type 98 len 28
>    <<< Hangs for 30 seconds, then continues >>>
>    -
>    -
>    <<< Output from ls command >>>
>    I, [2011-11-16T11:55:11.549000 #1213352018]  INFO --
> net.ssh.connection.session[7de]: closing remaining channels (0 open)
>
> If I switch to Java 5 on AIX instead, the program runs without problems.
> Also, I can run the same program without problems on Windows XP and Java
> 6 (from Oracle).
>
> Any help in solving this problem would be much appreciated.

Hmm...  The first thing that comes to mind is "reverse lookups".  see
http://rubydoc.info/stdlib/socket/1.9.3/IPSocket#addr-instance_method

You could try strace or truss to see what is different in socket
handling between the two versions of Java which show a difference.

Kind regards

robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/