A rolling upgrade of a cluster is accomplished through an ordered series of stages. These stages are represented in Table 25-2.
Rolling Upgrade Stage | ||||
---|---|---|---|---|
Stage | Description of Tasks to be Performed | Runlevel | Member | Time |
Preparation | Backup the Cluster. | single-user or multi-user | Any | Time can vary greatly according size and number of file systems that are required to be backed up. |
Choose a Lead Member for Rolling Upgrade. | multi-user | |||
Undate Firmware. | console | lead | ||
#clu_upgrade-vc check setup 1 | multi-user | |||
setup | #clu_upgrade -v check setup 1 | multi-user | lead | 45 - 120 minutes |
Reboot all cluster members eccept the lead member. | All but the Lead | |||
Preinstall | #clu_upgrade preinstall | multi-user | Lead | 15 - 30 minutes |
Install | #clu_upgrade install | multi-user | Lead | 30 - 120 mintues |
Update Installation (installupdate) and/or | single-user | |||
New Hardware Delivery (NHD) kit installation (nhd_install) and/or | single-user | |||
Patch kit installation (dupatch) | single-user or multi-user | |||
Postintall | # clu_upgrade postinstall | multi-user | Lead | <1 minute |
Roll | one member at a time, shutdown the system. | console | All but the Lead | 15 - 30 minutes per cluster memnber |
Update the firware. | ||||
Boot to single-user mode. | ||||
while in single-user mode #bcheckrc #update #swapon -a #clu_upgrade roll | single-user | |||
Switch | On any member, run "clu_upgrade switch" | multi-user | Any | <1 minute |
Reboot all members one at a time. | All | Normal reboot time. | ||
Clean | On any member, run "clu_upgrade clean" | multi-user | Any | 30 - 90 minutes |
Undo | Prior to the switch step, any or all steps can be undone. See section 25.6 for more information. | multi-user | varies according step being undone. | varies according step being undone. |
The times specified in this table are approximate and depend on the speed of the system, the speed of the storage, and whether or not the command is being executed on the CFS server (for the cluster_root (or /), cluster_usr or (/usr), and cluster_var (or /var) file systems). Please also note that the time required for the Setup Stage can be quite lengthy (beyond 120 minutes) – depending on the number of nodes in your cluster. The reason for this is that copies of all system level files are made on disk.