The following Q&A was compiled from “Ask the Expert: EMC storage best practice for upgrade and maintenance (Chinese Version)”.

Question 1: When upgrading the VNX operation environment, I can’t be absolutely certain what are the primary risks? What should be considered?

Answer:

In general, the customer is very careful with the risk of upgrade. EMC has a very strict process to control the risk of upgrade. Below are the whole process how RCM team (Remote Control Management Team) manage your upgrade risk:

1. The PCE of RCM team will make a preliminary check before upgrading the VNX operation system. If the box doesn’t pass the pre-check, we will notify the customer to fix this problem with EMC technical support center.

2. The RCM team will verify whether there is any known bug or compatible issue before upgrading the VNX operation system.

3. The upgrade engineer will make a double check before the same day of system upgrade.

4. The upgrade engineer will check system by running the relevant upgrade script.

5. The upgrade engineer will monitor the upgrade process.

6. After completing the upgrade, we will do a test and compare the result. If there is any issue, we will resolve it.

There is a need to pay particular attention that the upgrade of VNX-File have a system interruption about 3 minutes. In addition, the upgrade of VNX-Block doesn’t have service interruption. Our engineers will notify the customer what will happen before the system upgrade.

Question 2: Why don’t some customers want to upgrade to the latest version?

Answer:

Each new version can fix some bugs, however, however, there are some other reasons that customer do not want to upgrade maintenance.

Question 3:

When I make performance analysis, which indicators should I use? What is the corresponding solution? In the addition, how many IOPS can VPLEX with single engine support? If the back-end storage has many Flash drives, using VPLEX do mirror, whether VPLEX is the performance bottleneck? 

Answer:

When doing performance analysis, you should focus on the following indicators:

•     Utilization

•     Queue Length

•     Response Time

•     Total Bandwidth

•     Total Throughput

•     Read / Write Bandwidth

•     Read / Write Size

•     Read / Write Throughput

•     Service Time

Combined with customer service requirements for the corresponding optimization, if the customer has the relevant requirements, you can use the relevant EMC tools (Analyzer) to analyze the specific optimization to specific analysis. The system should use the safest and stable version.

VPLEX with single-engine can load 400000 IOPS in the lab, so we rarely observe VPLEX is the performance bottlenecks. If VPLEX is overloaded, the CPU utilization of VPLEX will become very high. If the CPU utilization of VPLEX is very high, we need to look for the root reason on a case-by-case basis.   Adding engine is one of the solutions.

Question 4:

The customer has an AX4 storage, and then he need to add six 1TB hard drives to the storage. What common types of risk are there? Are there any relevant document about AX4 disk expansion and DAE expansion?

Answer:

There isn’t any risk in the AX4 disk expansion. As I know, there are some bugs in the old version of AX code, you can refer to KB3112. About AX4 expansion document, you are able to generate it by EMC SolVe-Desktop tool.

Question 5:

Could you please introduce the OneFS operation system upgrade? Is there any impact for front-end IO when upgrading process?  Can we have two different version of the node in the same cluster?

Answer:

There are two options available for upgrading the OneFS operation systems: a simultaneous upgrade or a rolling upgrade.

In general, simultaneous upgrade is able to upgrade directly to the target version. However, OneFS 6.5.5.x or earlier versions aren’t able to directly upgrade to 7.2 versions. We may be required to upgrade to 7.0 or 7.1 versions, and then upgrade to 7.2 versions. A simultaneous upgrade require a temporary interruption of service during the upgrade process, all nodes must be rebooted.

For a rolling upgrade, you should verify whether EMC Isilon cluster is compatible with the new version of OneFS before upgrading the OneFS operation system. You can refer to the following document: https://support.emc.com/docu46145_Current-Isilon-Software-Releases.pdf. By default, NFS client is configured to auto connect, when the node is rebooted, NFS will automatically re-establish to other nodes. For SMB client, all SMB clients should manually connect to the cluster after node is rebooted. Although EMC don’t recommend the customer configure SAN in Isilon, some customers still configure iSCSI and VMware Datastore in Isilon, these data will be affected during the upgrade process.

Finally, one Isilon cluster doesn’t allow two different OneFS OS version.   

                                                                                                                                              

Translator: Jeffey