Upgrade Dell PowerFlex to v3.6.x

Preparing to upgrade physical Linux-based configurations

Steps to perform before upgrading PowerFlex on physical Linux-based nodes.

1. Install one of the following on the PowerFlex Gateway and PowerFlex presentation server nodes, as described in detail in the system requirements in the Getting to Know Dell PowerFlex Guide:

   Option	Example
   Oracle JRE 1.8 build 301	Linux: rpm -i jre-8uXX-linux-x64.rpm Windows (for the PowerFlex Gateway only): Double-click the file: jre-8uXX-windows-x64.exe
   OpenJDK JDK 8 build 302.b08 or OpenJDK 11.0.12.0.7 NOTE:JAVA 8 is not supported on PowerFlex v3.6.3 and later versions. For more information, see Dell PowerFlex 3.6.2 Release Notes.	Refer to the following information: Linux: https://jdk.java.net/archive/ Windows (for the PowerFlex Gateway only): https://github.com/ojdkbuild/ojdkbuild#downloads-for-windows-x86_64

2. Download and extract the PowerFlex packages.
   You need the packages for your base version and for the latest version (the version to which you are upgrading).
3. If you plan to use a PowerFlex Gateway that did not initially deploy your PowerFlex system, add the IP addresses of the system's primary and secondary MDMs to the gatewayUser.properties file, mdm.ip.addresses property, separated by semicolons. Both management and data IP addresses must be listed.
   Use a text editor to open the gatewayUser.properties file, located in the following directory on the PowerFlex Gateway node:

   PowerFlex Gateway operating system	Location of gatewayUser.properties file
   Windows	C:\Program Files\EMC\ScaleIO\Gateway\webapps\ROOT\WEB-INF\classes\
   Linux	/opt/emc/scaleio/gateway/webapps/ROOT/WEB-INF/classes

4. For data networks using IPv6, if you plan to implement a floating virtual IP address for the MDM, disable the IPv6 DAD setting on the node's relevant interfaces, using the command:

    sysctl net.ipv6.conf.<interface_name>.dad_transmits=0

5. If your SDSs contain NVDIMMs, ensure that the ndctl and daxio utilities are installed on each node before you commence the upgrade.
   • ndctl package—for example: ndctl-64.1-2.el7.x86_64.rpm
   • daxio package—for example: daxio-1.5.1-2.1.el7.x86_64.rpm

   CAUTION:Failure to meet this requirement could cause data loss.
6. If your SDSs contain NVDIMMs, spare capacity equivalent to the capacity of the NVDIMMs is required during the upgrade process. Ensure that sufficient capacity is available in the /tmp directory of each SDS. If you would like to use a different directory, you can configure this in the gatewayUser.properties file.
   1. Using a text editor, open the gatewayUser.properties file, located in the following directory on the PowerFlex Gateway node:

      PowerFlex Gateway operating system	Location of gatewayUser.properties file
      Windows	C:\Program Files\EMC\ScaleIO\Gateway\webapps\ROOT\WEB-INF\classes\
      Linux	/opt/emc/scaleio/gateway/webapps/ROOT/WEB-INF/classes

   2. Add the following line to the file:
      nvdimm.granularity.format.backup.path=<PATH>

      where <PATH> is the path to the directory that will be used for NVDIMM backup.

   3. Save and close the file.
   4. Restart the PowerFlex Gateway service:
      • Windows: From the Windows Services window, restart the EMC ScaleIO Gateway.
      • Linux: Type the following command:

        service scaleio-gateway restart

7. If you intend to use the CHAP component authentication feature after the upgrade, make the necessary preparations to ensure that you upgrade all system components now. Systems containing any components running PowerFlex versions lower than v3.5 do not support CHAP authentication.
   NOTE:Upgrade to v3.5 sets the CHAP authentication feature to DISABLED, and passwords will be all-zero. To enable the feature, after the upgrade is complete on all system components, follow the post-installation procedures described in the Configure and Customize Dell PowerFlex.
8. In SDSs containing NVDIMMs, verify that there are no bad blocks on the NVDIMMs. Bad blocks might prevent upgrade of the corresponding SDS. Bad blocks might be present when an NVDIMM's status is Normal. If there are bad blocks, replace the NVDIMMs with bad blocks before upgrading the node.
   1. Using shell, run the following command on each Linux-based node that has an NVDIMM:

      ndctl list --media-errors

   2. If the output for any of the hosts displays a result for "badblock_count":, one or more NVDIMMs (devdax device) have bad blocks. For example:

      {
       "dev":"namespace1.0",
       "mode":"devdax",
       "map":"dev",
       "size":62914560,
       "uuid":"47072fee-1447-4e04-b4ca-f28e2c03af80",
       "chardev":"dax1.0",
       "align":4096,
       "badblock_count":13,
      	"badblocks":
       }

   3. Refer to the Linux NVDIMM replacement procedure for the specific server model.