In one of my previous article I have mentioned that VPLEX upgrade as advertised by EMC is not completely online/non disruptive upgrade.
We have done 3 VPLEX code upgrade on past 2 months and we have come across many facts and requirements which are actually not written in any document. When we talk about VPLEX upgrade first thing EMC says is its an online non-disruptive upgrade but after that they open the pandora box of clause which need to be met to make the upgrade non-disruptive. All these clause are never mentioned during the install or configuration of the VPLEX. Which initial configuration these clause are not effective. This is the main reason of writing this article as not all the required pre-requisites and its remediation is online. So if it is not online then it means lots of planning, lots of man hours and lots of approvals.
Here are the EMC VPLEX upgrades pre-requisites which we have found till now. All these need to be fulfilled to make sure that the upgrade complete without outage.
1: LUN 0 for Storage view: All the Storage view in VPLEX config should have one LUN with HLU 0. if this condition is not met the server can face disconnection issue during upgrade due to Front end port high usage.
2: AIX ODM settings: Although this is pre-requisite for migration to VPLEX but if this condition is not met before the VPLEX code upgrade the server will face issue. So this change is much and it outage on the AIX server to make ODM setting changes. It is applicable only to AIX servers, AIX HBA parameters: fast_fail and dyntrk are both enabled. VPLEX NDU (Non-Disruptive Upgrade) causes Data Unavailable (DU) on AIX host.
Change the Reset_delay value from 2 to 0 in ODM
3: Only 4 Active path from backend: For all the devices attached from the backend array there should be only 4 active path, if the active paths are more than 4 VPLEX will throw error during pre-check for upgrade.
4: LINUX multi-path config: For Linux OS version less that 5 the multi-path setting in the multipath.conf file is must. Without this setting the server will loose connection to SAN during upgrade.
5: Meta volumes: When we have multiple SAN array attached to the back-end of VPLEX. one Meta volume need to be created using both the array to make sure redundancy is there for the array database.
6: Back-end array limitations: If we wanted do an upgrade on VNXD , VPLEX Geo synchrony version should be at least 5.3 SP1 otherwise it may experience an outage while doing code upgrade on VNX.
7: Firmware code file backup: If you are doing VPLEX code upgrade with multiple jump (you doing multiple level upgrade in code) please make sure you keep a copy of code file safe with you separately as during the each level of upgrade the VPLEX will delete the code loaded during upgrade and you need to copy the file again to the VPLEX to move ahead with the upgrade.
8: Connectivity from Multiple FE directors: Each storage view must at least have 2 paths from different directors & recommended is to have from all available directors.
These are few points which we faced while working on VPLEX code upgrades. Hope these help others too while doing upgrade in their environment.
Credits to my friend Saran for helping with this.