About EOS12.0 Fast Upgrade

[shuhei]

Did you remove the process that was previously in place to save time?

(Did you shorten the wait time?)

Also, is the reduction time about 50%?

=============================

Faster upgrades

Element 12.0 brings decreases in upgrade times.

Upgrades from Element 12.0 to future Elementreleases are significantly faster than upgrades from previous versions.

https://library.netapp.com/ecm/ecm_download_file/ECMLP2865022

=============================

[NetApp_AU · Accepted solution]

Hello shuhei,

The 12.0 upgrade process still works the same as previous releases.

The time for each node to install the update and reboot has been improved. I am not sure by what percentage the upgrade is faster though.

Note: If you have not upgraded to 11.8 first and performed the full power cycle on H610S nodes, you can still upgrade to 12.0, but you have to perform the full power cycle on the H610S nodes after the 12.0 upgrade. This will add time to the upgrade.

If you have upgraded to 11.8 first and performed a full power cycle already, then you do not have to perform the full power cycle again for the 12.0 upgrade.

See documentation below for more details.
https://docs.netapp.com/sfe-118/topic/com.netapp.doc.sfe-ug/GUID-823B29AE-E6F3-4DC3-97D3-7DFA2954139F.html

https://docs.netapp.com/us-en/hci/docs/task_hcc_upgrade_element_software.html#upgrading-h610s-storage-nodes-to-element-12-2-phase-2

https://kb.netapp.com/Advice_and_Troubleshooting/Hybrid_Cloud_Infrastructure/H_Series/NetApp_H610S_storage_node_power_off_and_on_procedure

[shuhei]

Thank you for the information.

I have checked internally and found that the following parts have been changed in time.

-With the new RTFI (uRTFI), the backup is no longer performed before the upgrade, but instead is rolled back.

-Parallel Threading in Cluster Failure Checking.

-Improved cache flushes on upgrades.

We also saw in the ActiveIQ Event log that the following times.

Depending on the number of iSCSI sessions, it could take longer.

24 node production cluster went from 26 hours in 10.x to 11 hours in 11.x, will be 6 7 hours in 12.x.