Symptoms

A partition's configuration can get wiped out and reset to default when the controller is upgraded to VELOS 1.6.4, v1.8.1, or v1.8.2.

Impact

-- The partition configuration data is wiped and reset back to default.

Conditions

If a partition was created while the system controller was running VELOS v1.1.x, the partition configuration volume may have been created with as an LVM "thin pool" rather than a logical volume. The problem does not occur with partitions created on VELOS 1.2.0 or later. In VELOS v1.6.4, v1.8.1, and v1.8.2, the partition startup logic incorrectly fails to recognize that these filesystem volumes are already initialized, and reinitializes them resulting in data loss. The reinitialization error is persistent, and will be repeated any time the partition instance is restarted, until the volumes are recreated without "thin provisioning".

Workaround

Before upgrading to the system controllers to VELOS v1.6.4, v1.8.1, or v1.8.2, check if the partition logical volumes on either system controller are incorrectly provisioned. To do this, run the "lvs" command from a bash shell on both controllers. "Thin provisioned" volumes will have numeric entries in the "Data%" and "Meta%" columns, and will have the "t" flag in the "Attr" column In this example, partitions 1 is vulnerable: [root@controller-1:Active log]# lvs LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert partition1 partition_config twi-a-tz-- 10.00g 0.00 10.61 partition3 partition_config -wi-ao-z-- 10.00g export_chassis partition_image -wi-ao---- 256.12g partition1 partition_image twi-a-tz-- 15.00g 0.00 10.57 partition1_shared partition_image -wi-a----- 10.00g partition3 partition_image -wi-ao-z-- 15.00g partition3_shared partition_image -wi-ao---- 10.00g ... If this condition exists, do not upgrade to these versions and contact support for assistance for non-destructively converting these volumes.

Fix Information

None

Guides & references

K10134038: F5 Bug Tracker Filter Names and Tips