Monday, January 27, 2014

Timeout error during OpenStack installation

While trying to install OpenStack using RDO, following the instructions in the link here , you might face the following timeout error

Applying <IP address>_prescript.pp   [ ERROR ]

ERROR : Error appeared during Puppet run: <IP address>_prescript.pp
Error: Command exceeded timeout

The reason for this error, in all probabilities would be the delay in your network, which could be causing the netns component packages to be timed out You can increase the timeout by editing the following file


Add a timeout line in the file

    if $::netns_support != "true" {
        exec { "netns_dependecy_install":
            path => "/usr/bin/",
            command => "yum update -y kernel iputils iproute",
            timeout => 1900,

Now you can retry the installation. It is advisable to use the answer file that is created during the initial installation, otherwise the reinstall would fail

Sample command for doing the reinstall using answer file is

packstack --answer-file=/root/packstack-answers-20140125-195230.txt

replace the text file name with the name of your answer file that was created during the initial installation

Thursday, January 9, 2014

XenCenter troubleshooting

Issue 1:

While trying to add a new host to a Pool in XenCenter, I was getting error "internal connection failed no route to host "


It was quite simple actually, but since I couldnt find it mentioned in any knowledge bases or articles, I am noting it down here. The machine where XenCenter was installed couldnt resolve the Xen host name. The machines were not added to domains and hence no DNS entries were there. So I had to add a host entry to the XenCenter machine to help with the name resolution

Issue 2:

Enabling AD authentication via XenCenter fails with error "clock skew detected with active directory server"


This happens because of the time difference between XenServer and AD server


Update NTP server list of the XenServer and point it to the AD server of the domain. Steps below

From the XenServer console, go to 'Network and management interface"->Network time(NTP)

Select option "Remove all NTP servers" to remove the default NTP server configuration

Then select option "Add an NTP server"

Enter the name or IP address of the NTP server

This will update the NTP configuration. On the same menu, click on 'Enable NTP time synchronization" to complete the configuration


Wednesday, January 8, 2014

Xen learnings

This week, I was trying to get my head around a new Hypervisor(new for me, obiviously), ie XenServer. Though it is pretty much same as ESXi and is free as well, there are some striking differences as well.The observations are based on the free version of Citrix XenServer version 6.2.0

- While ESXi needs a paid vCenter to manage multiple hosts, you can use the free XenCenter software to manage multiple Xen hosts

- Latest version of Xen server doesnt have the equalent of DRS or DPM. There was a feature named workload balancing, which was strangely discontinued in version 6.2.0 stating reason that there are no takers for it.

- It does offer a High Availability of VMs using pool based clustering of hosts

-XenMotion is the equalent of VM live migration, but it is restricted to one VM at a time

-XenCenter doesnt have a web client like vCenter

-There was a tool named XenConvert used for physical to virtual conversion, but it is retired as well.

-There is an option named Dynamic Memory Control(DMC), which can be used for dynamic allocation of memory for VMs. We can set a maximum and minimum memory options for VMs which will be used by XenServer to manage memory crunch situations

-Thin provisioning is supported  for local storages only

-Distributed vSwitch controller appliance  is available for centralized management of networks in XenCenter. However, this too is being depreciated in v6.2.0

PS: One interesting point to note is that the configuration limits document of xenServer is very small when compared to VMware and it doesnt have much details mentioned. For example, VMware specified the maximum number of vCPUs that we can create per physical processor core. For v5.1 its 25 and for v5.5 its 32. However Xen doesnt give you a hardcoded value for that. When we contacted Xen support regarding the same, they mentioned that there is no limit!!! Obiviously, that means you have to keep performance in mind while deciding on the number of vCPUs

Here is a good comparison matrix between various hypervisors available in market