Thanks Aaron.

It looks like we're in good shape as far as this stuff is concerned...

 $ /usr/local/vcl/bin/perl -MRPC::XML -e 'print "success\n";'
 success

 $ /usr/local/vcl/bin/perl -MLWP -e 'print LWP->VERSION ."\n";'
 6.04

...but I wonder if it's possible that under some circumstances vcld is
calling a perl other than the one we purpose-installed for it.  vcld
itself was adjusted to call that instance but a couple of other perl
scripts in vcl/bin weren't, so I've fixed that and scheduled another
reservation 45 minutes from now, will see if that makes any difference.



On Thu, Oct 04, 2012 at 01:42:47PM +0000, Aaron Coburn wrote:
>    I have seen something like this before.
> 
>    I would recommend ensuring that the RPC::XML perl module is properly
>    installed on the management node. Try this command:
> 
>    perl -MRPC::XML -e 'print "success\n";'
> 
>    If you get errors from this, try manually installing the RPC::XML
>    module from here:
> 
>    [1]http://search.cpan.org/~rjray/RPC-XML-0.77/
> 
>    The only thing to note is that the current version of RPC::XML requires
>    at least version 5.834 of LWP. To check which version of LWP you
>    currently have installed, run this command:
> 
>    perl -MLWP -e 'print LWP->VERSION ."\n";'
> 
>    If this reports any version before 5.834 (as my system does), then you
>    may need to use version 0.72 of the RPC::XML module, which requires
>    only version 5.801 of LWP.
> 
>    Also, when I say 'manually install', I do not mean using the cpan
>    interactive interface; rather, download the package with wget, and then
>    use the standard perl installation sequence:
> 
>    perl Makefile.PL
> 
>    make
> 
>    make test
> 
>    make install
> 
>    Aaron
> 
>    --
>    Aaron Coburn
>    Systems Administrator and Programmer
>    Academic Technology Services, Amherst College
>    [2][email protected]
>    On Oct 4, 2012, at 9:05 AM, Michael Jinks <[3][email protected]>
>    wrote:
> 
>      Okay, this makes sense.  Thanks for clearing that up; I'd wondered
>      if
>      vcld had logic for ramping up to a reservation.
>      I made a reservation last night, scheduled for 08:00 this morning.
>      At
>      02:00, I have errors in my log, which I'll paste below.  I haven't
>      started trying to investigate what's wrong; the error suggests a bad
>      cert.  Our cert is signed by Comodo, maybe that's the problem?  I
>      don't
>      know what vcld uses as a trust chain.
>      There could be several other things going on too.  The reference to
>      'Can't locate object method "type"' looks ugly, but I wonder if it's
>      just a result of the SSL error further up?
>      Log excerpt follows.
>      [...]
>      2012-10-04
>      02:00:02|3322|blockrequest|blockrequest.pm:process(192)|processing
>      blocktime_id= 15 pass 1
>      2012-10-04
>      02:00:02|3322|blockrequest|utils.pm:xmlrpc_call(9105)|argument_strin
>      g= XMLRPCprocessBlockTime 15 1
>      Can't locate object method "type" via package
>      "RPC::XML::Client::send_request:
>             HTTP server error: Can't connect to [4]vlab.uchicago.edu:443
>      (certificate verify failed)" (perhaps you forgot to load
>      "RPC::XML::Client::send_request: HTTP server error: Can't connect to
>      [5]vlab.uchicago.edu:443 (certificate verify failed)"?) at
>      /usr/local/vcl-2.2.1/bin/../lib/VCL/utils.pm line 9121 (#1)
>      Uncaught exception from user code:
>             Can't locate object method "type" via package
>      "RPC::XML::Client::send_request: HTTP server error: Can't connect to
>      [6]vlab.uchicago.edu:443 (certificate verify failed)" (perhaps you
>      forgot to load "RPC::XML::Client::send_request: HTTP server error:
>      Can't connect to [7]vlab.uchicago.edu:443 (certificate verify
>      failed)"?) at /usr/local/vcl-2.2.1/bin/../lib/VCL/utils.pm line
>      9121.
>      at /usr/local/vcl-2.2.1/bin/../lib/VCL/utils.pm line 9121
>             VCL::utils::xmlrpc_call('XMLRPCprocessBlockTime', 15, 1)
>      called at /usr/local/vcl-2.2.1/bin/../lib/VCL/blockrequest.pm line
>      373
>             VCL::blockrequest::process_block_time(15) called at
>      /usr/local/vcl-2.2.1/bin/../lib/VCL/blockrequest.pm line 193
>             VCL::blockrequest::process('VCL::blockrequest=HASH(0x358bb68)
>      ') called at /usr/local/vcl/bin/vcld line 568
>             VCL::vcld::make_new_child('HASH(0x34740f0)') called at
>      /usr/local/vcl/bin/vcld line 448
>             VCL::vcld::main() called at /usr/local/vcl/bin/vcld line 98
>      2012-10-04
>      02:00:02|3322|blockrequest|State.pm:DESTROY(829)|VCL::blockrequest
>      destructor called, address: 358bb68
>      2012-10-04 02:00:02|3322|blockrequest|State.pm:DESTROY(848)|number
>      of database handles state process created: 1
>      2012-10-04 02:00:02|20597|vcld:REAPER(718)|VCL process exited for
>      reservation <unknown>, PID: 3322, signal: CHLD
>      On Thu, Oct 04, 2012 at 08:48:34AM -0400, Aaron Peeler wrote:
> 
>      Hello Michael,
>      The start time for block allocations need to be requested > 30
>      minutes
>      in advance.
>      If folks think this should be adjusted let us know. The original
>      thinking was to allow the enough time for the machines to get
>      prepared
>      for 30+ machines, which is what we've experienced for the average
>      number of machines for Blocks allocations.
>      Aaron
>      On Wed, Oct 3, 2012 at 6:03 PM, Michael Jinks
>      <[8][email protected]> wrote:
> 
>      Thanks, Aaron and Josh.
>      I did have some problems in my vcld.conf; the vclsystem account
>      hadn't
>      been properly configured.  So that was clearly part of my trouble.
>      Having fixed that, though, I'm still not getting a successful block
>      reservation.  Based on Josh's remarks about vcld reserving the slots
>      "several hours" in advance, I wonder if that's an issue.  In order
>      to
>      test this in a reasonable amount of time, I've been creating a time
>      slot
>      a few minutes in the future, which then comes and goes with no
>      activity
>      in the vcld log except 'lastcheckin' lines.  I also find, in the
>      database:
>      mysql> select processed from blockTimes;
>      +-----------+
>      | processed |
>      +-----------+
>      |         0 |
>      |         0 |
>      +-----------+
>      2 rows in set (0.00 sec)
>      How far advance do I need to do a block allocation in order for it
>      to
>      take effect?
>      On Wed, Oct 03, 2012 at 03:03:46PM -0400, Josh Thompson wrote:
> 
>      -----BEGIN PGP SIGNED MESSAGE-----
>      Hash: SHA1
>      Michael,
>      Aaron gave some good info.
>      One thing I'll point out is that when you view the status of a block
>      allocation through the web interface as you've described, the number
>      listed for "Failed" comes from
>      (number of computers for block allocation) - (number of computers in
>      blockComputers table)
>      It doesn't necessarily reflect that VCL attempted to load the
>      machines
>      and that those loads failed.  If vcld never processes the block
>      allocation, it will always show the full count in the Failed row.
>      When the block allocation is created, an entry is created in the
>      blockTimes table for each time slot in which the block allocation
>      will
>      occur.  However, no computers are actually allocated for each time
>      slot until several hours before a time slot starts.  Several hours
>      before any block time, vcld processes that block time to allocate
>      the
>      computers and populate the blockComputers table for that blockTimes
>      entry.  It does this by making an XMLRPC call to the web code.  You
>      can check that vcld has actually processed a block time entry by
>      looking at the blockTimes.processed field in the database for a
>      given
>      time slot.
>      Josh
>      On 10/03/12 13:59, Aaron Coburn wrote:
> 
>      Michael,
>      You should make sure that /etc/vcl/vcld.conf has the proper values
>      set for:
>      xmlrpc_username xmlrpc_pass xmlrpc_url
>      xmlrpc_username should be set as vclsystem@Local xmlrpc_url should
>      be something like:
>      [9]https://vcl.myinstitution.edu/index.php?mode=xmlrpccall
>      This is all described in the vcld.conf file as well as on this
>      page:
>      https://cwiki.apache.org/confluence/display/VCL/VCL+2.3+Management+N
>      ode+Installation#VCL2.3ManagementNodeInstallation-Configurevcld.conf
>      Then be sure to set the password for the user identified above
>      (e.g. vclsystem@Local). Instructions are here:
>      https://cwiki.apache.org/confluence/display/VCL/VCL+2.3+Management+N
>      ode+Installation#VCL2.3ManagementNodeInstallation-Setthevclsystemacc
>      ountpasswordforxmlrpcapi
>      (The instructions are for version 2.3, but they will work for
>      2.2.1 as well)
>      Finally, restart vcld so that it can pick up the new information.
>      If that doesn't help, then try inspecting the vcld logfile -- it
>      will tell you more about why the reservations fail.
>      Aaron
>      -- Aaron Coburn Systems Administrator and Programmer Academic
>      Technology Services, Amherst College [email protected]
>      On Oct 3, 2012, at 1:30 PM, Michael Jinks <[email protected]>
>      wrote:
> 
>      Hi List.
>      We have a VCL 2.2 instance running, most things work, but one
>      glaring problem is still thwarting me.  When I try to create a
>      block allocation, the computers always come up "Failed"; this in
>      spite of the fact that I can reserve the same image individually
>      with no issues.
>      I've tried this so far with the admin@Local account, as well as
>      my user account which has full admin privileges (or at least as
>      close as we can get by assigning rights in the Privileges page).
>      I go to the "Block Allocations" page, click "Create New Block
>      Allocation", fill in the form, choosing a known-working image in
>      the "Environment" dropdown, and a "List of Dates/Times" with one
>      time span in the near future.  "Submit New Block Allocation" pops
>      up a confirmation window, that goes fine, and I have an entry
>      "Block Allocations" page under "Manage Block Allocations".  Looks
>      fine.
>      But, if I click an entry under "Your Active Block Allocations",
>      under "Current status of computers," the number by "Failed" is
>      equal to the number of seats I reserved, and as far as I can tell
>      the system never tries to deploy a VM.
>      I'm guessing that there is some permission setting or some other
>      item within the management node that I've overlooked.  Any hints
>      on where to look?
>      Thanks, -j
>      -- Michael Jinks :: [email protected] University of Chicago IT
>      Services
> 
>      - --
>      - -------------------------------
>      Josh Thompson
>      VCL Developer
>      North Carolina State University
>      my GPG/PGP key can be found at [10]pgp.mit.edu
>      All electronic mail messages in connection with State business which
>      are sent to or received by this account are subject to the NC Public
>      Records Law and may be disclosed to third parties.
>      -----BEGIN PGP SIGNATURE-----
>      Version: GnuPG v2.0.17 (GNU/Linux)
>      iEYEARECAAYFAlBsjBIACgkQV/LQcNdtPQOcyQCfTQE1TzGXVFsIqitpPFrzJrM/
>      SoIAn3p/PMlH+mjI0jWpHdwOC12/hwyu
>      =gnM2
>      -----END PGP SIGNATURE-----
> 
>      --
>      Michael Jinks :: [11][email protected] :: 773-469-9688
>      University of Chicago IT Services
> 
>      --
>      Aaron Peeler
>      Program Manager
>      Virtual Computing Lab
>      NC State University
>      All electronic mail messages in connection with State business which
>      are sent to or received by this account are subject to the NC Public
>      Records Law and may be disclosed to third parties.
> 
>      --
>      Michael Jinks :: [12][email protected] :: 773-469-9688
>      University of Chicago IT Services
> 
> References
> 
>    1. http://search.cpan.org/~rjray/RPC-XML-0.77/
>    2. mailto:[email protected]
>    3. mailto:[email protected]
>    4. http://vlab.uchicago.edu/
>    5. http://vlab.uchicago.edu/
>    6. http://vlab.uchicago.edu/
>    7. http://vlab.uchicago.edu/
>    8. mailto:[email protected]
>    9. https://vcl.myinstitution.edu/index.php?mode=xmlrpccall
>   10. http://pgp.mit.edu/
>   11. mailto:[email protected]
>   12. mailto:[email protected]

-- 
Michael Jinks :: [email protected] :: 773-469-9688
University of Chicago IT Services

Reply via email to