Presented by Chris Colotti, Consulting Architect with VMware (blog, twitter) and David Hill, Senior Consultant with VMware.

Got into this session partway through so apologies for the incomplete notes.

Summary = vCD 1.0 to 1.5 is not for the faint of heart, can involve a good bit of downtime, and was a serious effort even for VMware staff with direct access to just about anyone inside VMware.

Update: so it turns out Chris has a four part series on this on his weblog….go here for part 1. Oddly enough, he’s just a wee bit more comprehensive than my notes below. 😉 (although my notes do emphasize what jumped out to me + some audience questions/responses). Read both basically…

More after the jump…

• Phase 1 Key Takeaways
  • Downtime is required – all cells must be shut down.
    • Mandatory for db upgrade.
    • Use Cell Management Tool to quiesce tools
  • Need time sync everywhere.
  • Chargeback should be stopped at various points.
  • Backup all db’s concurrently before each upgrade.
  • vCloud Connector 1.0 isn’t support with vCD 1.5
  • Use VMware snapshots on the vCloud Cell and vCenter management VMs before upgrade for rollback (vShield Manager)
  • Don’t deploy new vShield Manager .OVA – use the upgrade tar.
  • Upgrade chargeback last.
  • New license for vShield 5.0 is required – be sure to apply them!
  • Don’t delete existing OVA and setup new one…you’ll lose a very important MySQL database.
• Phase 1 Timeline – could be 2 hours in perfect world….but it took them 1-2 weeks realistically.
• Phase 2 Key Takeaways
  • Can use vCenter Heartbeat to fail over vCenter, upgrade one side, fail it back, upgrade other – vCloud director doesn’t go down at all.
  • Stop Chargeback services as it talks to vCenter
  • Stop vCloud cells during the vCenter upgrade.
  • 1.5 Cell Management tool commands are different than 1.0.01
    • Seemed to not return values w/using SSh but does with console.
    • Need the “cell” argument, switches are the same.
  • Follow all vCenter KB articles available for supported items.
  • Ensure all hosts are still showing “prepared” after upgrade.
    • 1.0 and 1.5 the agent installer is different. In 1.5 it’s DIB-based, 1.0 is the old one.
  • Custom roles created for service accounts gets updated.
  • New license for vSphere 5
  • “Refresh” required in vCloud Director
  • Cisco Nexus 1000 upgrade is unclear at this point.
• Phase 3 Key Takeaways
  • Requires no more downtime to the users.
  • Third-party agents may need to be removed before upgrading.
  • Use Update Manager if possible.
    • Will try to remove the 3rd party agents but sometimes doesn’t remove them all.
  • Can be done over time but will mean mixed environment.
    • Not supported permanently but is supported temporarily for upgrade.
• Phase 4 – Networking, Tools, Virtual Hardware.
  • Don’t do this phase until everything is ESXi 5.0
  • N1KV upgrade is unknown when to put into the process.
  • Tools before virtual hardware – same as always.
  • VMFS-5 upgrade doesn’t change block size – same as always.
  • DISABLE Storage DRS
  • Fast Provisioning per Org vDC – may need new design considerations.
  • vApp owners can update tools and hardware themselves.
  • Ensure the vApps in the catalog are also redeployed and upgraded.

Can’t go from Oracle to SQL during upgrade — if started out with Oracle, have to do a new cloud to get to SQL.

Asking if should avoid N1KV (my question) – Not a “bad” on N1KV but just didn’t do it. Also a lot of customers on vCD 1.0 didn’t use N1KV as N1KV didn’t support VCDNI.

Asking if can enable Storage DRS but manually (my question) – being very cautious so saying no but will check with Denneman and blog about it.