OneView Appliance Reboot Time

[shocko]

Guys, our appliance is running the latest version 1.20.01-0021397, Dec 12, 2014  of OneView for vSphere. We had to reboot the appliance a couple of times today to resolve an issue whereby it was seeing the virtual connects in a 'maintenance' state. Reboot takes 20 mins on an appliance with 10 Gb and 2 vCPUs of compute!

[ChrisLynch]

The reboot of an appliance can take "some" time.  What impacts the appliance startup is the number of managed and monitored devices, and how many Logical Interconnects are being managed.  Also, the underlying hypervisor disk that hosts the appliance virtual disk file can impact performance.  But waiting 10-20 minutes for an appliance to reboot is not an issue, especially since the infrastructure the appliance is managing can continue to operate until the appliance is back online.

[shocko]

Hi Chris, I would argue that this length of time to boot has an impact. We were running a firmware upgrade yesterday on a Gen9 C-Class enclosures. We could not manage ti using OneView as ti was coming up with 'Maintenance Mode'. Rebooting the OneView appliance resolved the issue but had to wait 25 mins. Only a single enclosure with 3 blades registered so number of devices not a concern and running with 2x vCPU and 10 Gb RAM. The appliance spikes the vCPU at 100 during boot. No other virtual machine son the hypervisor using any CPU. Insight Deployment appliance behaves the same way. 

[ChrisLynch]

Are you referring to the VC Interconnects were coming up in Maintenance state during a firmware update?  Did the firmware update Task complete, or was it still running?  Because that is by design, were the VC modules will reboot into Maintenance state and go through an Add process again (which is the action OneView takes to reconfigure the modules.)

Yes, the appliance is CPU bound during boot.  You can't add more vCPU at this time.  Also, disk IO can be greatly impacted, so making sure the VMDK/VHDX is hosted on a fast enough disk/logical disk is pretty important.

[shocko]

Hi Chris. We needed to add several networks and upgrade firmware accross the enclosurse. We could not do this as the interconnects were showing in maintenance mode. The resolution was simply to reboot the appliance.  This takes ~ 25 mins, the longest boot in our environment! This meant we had to wait 25 mins before we could start work as you cannot get at the VC directly using SSH. As stated there is nothing pointing at OneView bar a few systems. I have even moved the appliance to a test ESXi node where no others vms are runing and the datastore is backed by SSDs and still the boot tile is 20m +. 

[ChrisLynch]

A reboot of an appliance should only be if other options have not resolved the particular issue you are experiencing.  Did you try any of the following:

• refresh of the Enclosure? 
• Reapply the Logical Interconnect Configuration?
• Reboot the VC interconnects 1 and at time?

Yes, the appliance reboot can take some time; what you are seeing isn't abnormal.  Rebooting an appliance should never be a common troubleshooting step.

[shocko]

Hi Chris, I'm not really concerned with the virtual connects etc. To say that we should never have to reboot the appliance is a bit of a long shot in my experience. All systems have leaks etc. and bugs that might require a reboot at some stage. We have 2 HP appliances here and they both take 25 mins to boot. My query is simply what might be causing them to take so long given that they have plenty of resources. From looking at the CPU counters in ESX, they spike on boot but then the appliance appears to load up mozilla and sit there for 15 mins with very little vCPU activity. This indicates the appliance it self  (or a process therin) seems to be waiting for something. 

[ChrisLynch]

I would argue you should have been concerned with the VC Interconnects and focused your time on why they were in Maintenance Mode.  Regardless if a reboot fixed it, a simple Refresh of the Enclosure should have done that anyhow, without waiting for the appliance to reboot.  Again, an appliance reboot should be the last case resort, and never part of any common troubleshooting steps.  There are cases where a reboot is necessary (being directed by HP Support and during an appliance software upgrade are two of the most common.)

[shocko]

Again, yes I am interested in the issue with VC, but I am not intersted in it in the context of this thread! See subject line and please stick to the question asked :) !

We have a call open with HP regarding the underlying issue with the VC going into maintenance mode in the first place. 

[shocko]

PS: the reboot was recommeded by the HP engineer onsite with us as he had seen this issue before. 

[Dennis Handly]

>All systems have leaks etc. and bugs that might require a reboot at some stage. 

Well, for HP-UX, that may be many many months.  :-)

Whereas applications running there may be a different matter.