You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@vcl.apache.org by Josh Thompson <jo...@ncsu.edu> on 2017/07/05 16:18:57 UTC

Re: Problem using multiple management nodes

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Joseph,

This was due to a bug in the initial database configuration.  I created a JIRA 
issue for it and fixed it.

The way to delete entries from privilege nodes is to just uncheck all of the 
assigned privileges/attributes.  The next reload of that node will show the 
item removed.  However, the "available" attribute is assigned for the 
allManagementNodes resource group, which as you pointed out, cannot be 
unchecked.  The following command should take care of it for you.  You'll 
probably need to log out/log in to see things correctly after running the 
command.

mysql vcl -e "DELETE FROM resourcepriv WHERE \`type\` = 'available' AND 
privnodeid = (SELECT id FROM privnode WHERE name = 'admin' AND parent = 
(SELECT id FROM privnode WHERE name = 'VCL')) AND resourcegroupid = (SELECT id 
FROM resourcegroup WHERE name = 'allManagementNodes');"

Josh

On Tuesday, June 27, 2017 7:33:13 PM Joseph Fischetti wrote:
> Thanks Andy,
> When I try to delete the "allManagemenetNodes" group, it doesn't let me
> because it's assigned to VCL > admin under privileges.  I'm not sure if
> it's just a problem with my interface, but there's no way to delete a
> resource from the privileges table.  allManagementNodes has no check marks,
> but I don't see an option to remove it. 
> Joseph Fischetti
> Linux System Administrator
> Marist College
>  
>  
> ----- Original message -----
> From: Andy Kurth <an...@ncsu.edu>
> To: user@vcl.apache.org
> Cc:
> Subject: Re: Problem using multiple management nodes
> Date: Tue, Jun 27, 2017 3:26 PM
>  
> Glad it's working.  You are correct.  The website checks the MN's last
> checkin time.  If it's too old, it assumes there's a problem and won't
> assign reservations to it. 
> Also, be sure to make sure each guest VM computer is only mapped to a single
> management node.  VM guests can be in multiple computer groups, but you
> wouldn't want any VM to be in multiple computer groups that are mapped to
> different management node groups.  You should make sure neither MN is are
> still a member of the allManagementNodes group and you can probably delete
> the allManagementNodes group to keep things clean. 
> -Andy
>  
> -Andy
>  
> On Tue, Jun 27, 2017 at 3:11 PM, Joseph Fischetti
> <Jo...@marist.edu> wrote: Following up for archival purposes:
> After looking through the database and logs I noticed the check-in time for
> IB-MN was a few hours off.  I corrected the time discrepancy and was able
> to create a reservation on IB successfully.  I had all permissions and
> settings correct. 
> I would imagine (devs can correct me if I'm wrong) that if the management
> node hasn't checked in recently, the front end automatically assumes it's
> unavailable, regardless of it's actual status.  Since it's last check in
> was 4 or 5 hours earlier, the front end didn't see an available management
> node and wouldn't assign a reservation for it. 
> Joseph Fischetti
> Linux System Administrator
> Marist College
>  
>  
> ----- Original message -----
> From: "Joseph Fischetti" <Jo...@marist.edu>
> To: user@vcl.apache.org
> Cc:
> Subject: Problem using multiple management nodes
> Date: Tue, Jun 27, 2017 11:51 AM
>  
> Hi,
> Sorry for the long winded question, but I'd like to be thorough:
>  
> We have a VCL with separate front end server, database server, and
> management node(s).  The compute resources that we would like available are
> on two separate boxes, for simplicity I'll refer to them here as IB, and SB
> (Ivy Bridge and Sandy Bridge respectively).  Both IB and SB are ESXi hosts
> with datastores and networking configured correctly, we'll assume. 
> Up until now, I had a single management node (SB-MN) running on SB.  The
> public IP for SB-MN is 10.10.5.201, the private IP for SB-MN is
> 172.20.105.201, and there's a DHCP server running on the two interfaces for
> handing out IP's from 210-219 for the 10 guests that the VCL handles on
> this box. We're in testing and will be scaling it up for production. 
> I'd like to make resources available on IB, so I cloned SB-MN to IB, and
> fixed the networking appropriately.  IB-MN public is 10.10.6.202 (don't
> ask), 172.20.106.202, and there's a DHCP server running on the two
> interfaces handing out IPs from 210-219. 
> I've created two new managementnode groups in the Resource groups area of
> the front end (SB-ManagementNode and IB-ManagementNode).  I've also created
> two new computer groups, SB-Guests, and IB-Guests.   
> I believe all grouping and mapping is correct, it follows the following idea
> for both SB and IB: IB-MN is in IB-ManagementNode group.
> IB-ManagementNode group maps to IB-Guests
> IB-Guests contains the 10 computers.
>  
> IB-ManagementNode and IB-Guests (as well as SB) are all part of the
> available resources in the privileges tables. 
> My problem is, with the above grouping/mapping, when I try to create a
> reservation I get the error that there is no time slot available.  It isn't
> until I put both management nodes in the "allmanagementnodes" group, and
> map the "allmanagementnodes" group to the guests that I can actually create
> a reservation.  Furthermore, if I continue on and create a reservation, the
> front end assigns that reservation to the wrong management node (i.e, only
> SB-MN is actually getting utilized).  I can confirm this by looking
> directly at the reservations table, managementnodeid always gets assigned
> to SB-MN 
> Truthfully, SB-MN can see/control the IB host, so the issue isn't with
> whether or not the guest machine is created.  The problem is that the
> private management networks for SB and IB are separate, so once SB-MN
> creates the guest on IB, it get's stuck waiting for 172.20.106.XXX to
> respond to SSH.  Which it never will. 
> I think there's an issue with how I have the management node groups
> configured, but I'm not sure.  Is there something in particular that needs
> to be done to "enable" another management node?  (both management nodes are
> set to available). 
> Joseph Fischetti
> Linux System Administrator
> Marist College
>  
>  
- -- 
- -------------------------------
Josh Thompson
VCL Developer
North Carolina State University

my GPG/PGP key can be found at 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-----

iEYEARECAAYFAlldEXEACgkQV/LQcNdtPQNWSgCfX32zzwlghbE2gYNm0G7UWaI5
41oAn2zhU3OXj6ySYjdGYb3eeAfla34f
=vYfy
-----END PGP SIGNATURE-----