Thursday Jul 25, 2013

Management access point authentication failure

Some users of Ops Center 12.1.4 are seeing an issue where, after upgrading or doing a new install, their assets are getting dozens of warning incidents that say "Management access point failed authentication during login to asset." Eventually, the severity increases as the warnings pile up, and communication with the asset can be disrupted.

You can prevent this issue, or halt it if you've already hit it, by logging in to the Enterprise Controller and creating a /var/tmp/no_os_selfhealing file if it doesn't exist. For example:

touch /var/tmp/no_os_selfhealing

You can use this fix before you do the upgrade or installation, to prevent the issue from occurring at all.

If you have assets that are already having communication trouble because of the issue, then once you've added the file as described above, you can solve the issue by deleting and rediscovering the asset deleting the Access Point for the Service Processor, and then rediscovering just the hardware. If you have a Proxy Controller or Enterprise Controller that's been affected, you should restart them after applying this fix.

Edit: You can also fix the issue by downloading the latest version of the OCDoctor (4.24) and running it with the --troubleshoot option. Take a look at this bulletin for more info.

Edit 2: Updated to use a less destructive fix for affected assets.

Tuesday Jul 23, 2013

Provisioning Different Versions of Oracle Solaris 11

I got a couple of questions about the new support for multiple Automated Installer (AI) releases in 12.1.4:

"How do I set up different versions of Oracle Solaris 11 in the Solaris 11 Software Library? For example, I want to use these versions for OSP:

  • Solaris 11/11 SRU 5.5
  • Solaris 11.1 SRU 8.4

So how do I get both of them into my library and ready for OSP use?"

There are a couple of ways. The first is to configure the parent repository to use a source with both of those releases. If you haven't initialized your library, configuring the parent repository is part of the library creation; otherwise it's a separate job, which you can set to repeat according to a schedule.

The other way is to use the Add Content wizard to add a specific S11 release to your library. This action doesn't save the parent repository location, but it works well if you're adding a specific version.

Once your S11 library has the content for both releases, you'll be able to create OSP profiles for either one. There's a dropdown option that lets you pick which version of S11 you want to use:

"I have my library using a valid repository, but the new versions aren't showing up in the OSP wizard. Do I need to restart the Enterprise Controller to make them show up?"

If the new versions aren't showing up, re-running the library sync job should make them appear. You shouldn't have to restart the EC.

Once you've done these steps, you can provision these versions of S11 from any of your Proxy Controllers.

Tuesday Jul 16, 2013

DHCP and OS Provisioning

I got a question about an OS provisioning error:

"When I create an Oracle Solaris 10 OSP profile, I get a "DHCP is not configured" error on the Select Networks step. What do I do about this?"

Well, it depends on what role you have. If you're an Ops Center Admin, you can go into Administration, select a Proxy Controller, and set DHCP up. There are two ways that you can do this - you can either configure DHCP on your Proxy Controllers:

Or, alternatively, if you have an existing DHCP server that you'd prefer to use, you can configure Ops Center to use it:

These tasks both require admin privileges, so you might need to bug your admin if you don't have that role yourself. But, either way, once you have DHCP set up you'll be able to do OS provisioning.

Tuesday Jul 09, 2013

Limits on LDom numbers

I got a question - or a few linked questions - about installing LDoms in a Server Pool through Ops Center.

I have some T5240s (2x SPARC T2+) and some T5440s (4x SPARC T2+) running as standalone Control Domains with some LDoms each.

1. Is there any limit to the number of CDoms per Server Pool within Ops Center?

No, there is not.

2. What about the number of LDoms per Server Pool, per CDom, or per server?

The only limitation on the number of LDoms is that a server or physical domain can't have more LDoms than it has threads, or have more than 128. If you get close to that limit and start having LDoms with one thread, performance will start to suffer.

3. Is it possible to mix them within the same server pool?

Yep. Depending on your LDom version and the hardware, live migration might not be possible, but you can have server pools with different hardware types, and automatic failover will work fine.

If you're also looking at setting up Server Pools through Ops Center, we put together a how-to about setting up a Server Pool which could be helpful. There's also a white paper on the OTN with best practices for setting up a virtualized SPARC environment.

Tuesday Jul 02, 2013

Migrating an LDOM from a T4 to a T5

I got a question about LDoms:

"Is there any restriction against migrating LDoms between the T4 and T5 platforms?"

The only restriction is that, at present, you can't do a live migration. However, with Ops Center 12.1.4, you can put T4 and T5s together in a Server Pool and either manually migrate the LDoms to a new host or configure them for automated cold-migration failover.

Take a look at the Server Pool and Oracle VM Server for SPARC chapters for more information.


This blog discusses issues encountered in Ops Center and highlights the ways in which the documentation can help you


« July 2013 »