Loss of configuration, policy or other vital system data after upgrade to SGOS 6.2.2, 6.2.3 or 220.127.116.11
Affected products and versions
SGOS 6.2.2.x and SGOS 18.104.22.168
A disk is at object store capacity when it has reached the maximum number of objects per disk it can hold. Systems that use the old disk layout, which is almost all systems in the field, have an object limit of 2.8 million objects per disk. A new disk layout introduced in 6.2 increases the object limit to varying values depending on the size of the disk. For example, a 500GB drive will support 13 million objects per disk and a 1TB drive will support 24 million objects per disk. Systems that use the new disk layout are not affected by this issue. When a system upgrades to 6.2 they do not automatically start using the new disk layout.
When a disk is at object store, an "overflow eviction" algorithm is used to best select an object to evict. The algorithm normally chooses objects with the old creation time but as an optimization would also preferentially choose objects whose creation time is prior to the cache reset time. This is because said objects cannot be served from cache, and thus they should be evicted first. A bug present in the affected products and versions introduced the issue where persistent system objects would be selected by the eviction algrythmn.
How to check if system could be affected by TFA72
This problem has been solved with code fixes for each affected line of SGOS.
Upgrade to a fixed SGOS release
SGOS 22.214.171.124 can be downloaded at:
WARNING: Performing a restore-defaults factory defaults on a 6.2+ build on a multi-disk system then the system will use the new disk layout by default. If the customer thinks they will need to downgrade to a pre-6.2 release, they should run the "disk decrease-object-limit" command on the CLI which will change the system to disk layout version 1.
Rate this Page
Please take a moment to complete this form to help us better serve you.