Advice on approaches to deploying servers at (modest) scale

[ChrisN2]

Hi,

I'm seeking general advice from the community from people with experience deploying HPE DL family servers at scale.

My organisation has grown from having 10’s of servers running ESXi and occasionally adding a few hosts manually to the point we will soon be heading to 130+ hosts and expanding in blocks of hosts 3 to 4 times a year. Prior to last year, all build, configuration and firmware updating was done manually. We have now implemented OneView to provide a level of central control. I find it does a good job with BIOS settings and firmware compliance but can’t set many of the iLO settings we generally apply. I’m hoping others in the community who have gone through this inflection point between manual and automated bulk provisioning or who run at significant scale can share advice or details of their approach.

Weak points with our current approach are:

• Getting iLO up and configured uniformly with a static IP, NTP, SNMP trap destinations etc.
• Network configuration for host ports. We have Arista CVP but no level of automated provisioning, port setup is manual.
• Getting ESXi deployed and patched (We’re still on 6.7 but I’m hoping the jump to 7 later this year and the introduction of lifecycle manager will help with this)
• Getting the ESXi configuration standardised. We use host profiles but there are limitations around datastores etc.

What should we be looking at as next steps to help us scale with less reliance on human intervention?

• DHCP on the iLO network to set as much as possible that way
• PowerShell scripting / XML based iLO config
• PXI boot and scripted installs of ESXi (or does vSphere 7 make this pain go away)
• Infrastructure as code solutions such as Terraform
• Something entirely different?

I’d love to hear any advice from people who’ve “been there, done that”!

Thanks

[ManBha]

Hello,

Hope the below document helps.

https://support.hpe.com/hpesc/public/docDisplay?docLocale=en_US&docId=a00115980en_us

https://techlibrary.hpe.com/docs/enterprise/servers/oneview5.0/cicf-api/en/index.html

Thanks.

[ChrisN2]

Hi ManBha,

Thanks for the OneView link but I'm familiar with it's capabilities. We set all the BIOS settings we need via server profile template and baseline the firmware but I find it lacking in the ability to set iLO configuration beyond adding a user and setting the hostname. Things I'd like to be able to set within iLO via profile are:

• DNS
• NTP
• SNMPv3 User
• SNMP trap destination
• Various security settings
• AD integration

Do you know if any of that is on the roadmap?

Thanks

[ManBha]

Hello,

Thank you for your prompt response.

Yes, going forward we might see these options available but not as of now.
When we add a server in OneView we consider that iLO is already configured and hence, these options might not be available to set via server profile.

Thanks.

If you feel this was helpful please click the KUDOS! thumb below and accept the solution.

[ChrisN2]

Hi ManBha,

I understand the thinking around iLO already being configured before importing into OneView. Do you have any good solutions I should look at to automate this intal configuration.

I'm aware there are REST/Powershell/command line ways of integrating with iLO and I've done some work on automating this step with PoweShell but I found it a little "brittle".

The ideal situation would be to connect the iLO dedicated NIC to a specific VLAN and have the iLO recognised, provisioned with the settings specific to that network (including issuing it a certificate) and added to OneView. Is there any off-the-shelf way of acheiving that?

Thanks