Changing the Cache Optimization Mode Incorrectly on Sun StorEdge 3310, 3510 or 3511 May Cause Issues Affecting Filesystem Availability and Data Integrity |
|
| Category : | AvailabilityData Loss |
| Release Phase : | Resolved |
| Product : | Sun StorageTek 3310 SCSI Array Sun StorageTek 3510 FC Array Sun StorageTek 3511 SATA Array
|
| Bug Id : | 5069625
|
| Date of Workaround Release : | 15-SEP-2004
|
| Date of Resolved Release : | 12-OCT-2005
|
Impact
Using sccli(1M) to change the cache optimization mode (sequential vs random) without first deleting all existing logical drives and then performing a reset of the array can affect filesystem availability and data integrity.
Contributing Factors
This issue can occur on the following platforms:
- Sun StorEdge 3310 (SCSI) Array without sccli 1.6.2
- Sun StorEdge 3510 FC Array without sccli 1.6.2
- Sun StorEdge 3511 FC Array (SATA) without sccli 1.6.2
The RAID controller operates with the same stripe size for all logical drives and the optimization mode is dependent on any existing logical drives configured. Should the operator change the optimization mode (i.e., "sequential" to "random"), without deleting all existing logical drives and resetting the array, the operating mode is still at "sequential" even though the optimization mode displays as "random." Any subsequent logical drives created will be labeled as "random" mode, even though the active operating mode is "sequential," and any data written to the newly created logical drives will be written using "sequential" mode stripe size instead.
The next time the controller(s) is reset or power cycled, the operating mode now changes to "random," as determined by the metadata on the existing logical drive(s), and a data mismatch will occur. Since the data was initially written as "sequential" and now the operating mode is "random," the data presented back to the host will appear to be out of sequence and corrupted as the stripe size and access pattern have been changed.
Symptoms
Data presented to host applications will appear to be corrupted and out of sequence. Volume management applications may not be able to import disk groups and fsck(1M) will be unsuccessful.
Workaround
The proper procedures to change optimization mode without running into a data mismatch are:
-
Back up data, if needed for later use
-
Delete all existing logical drives
-
Change optimization mode
-
Reset or power cycle the array unit
-
Create new logical drives
Resolution
This issue is resolved on the following platforms:
- Sun StorEdge 3310 (SCSI) Array with sccli 1.6.2 or later
- Sun StorEdge 3510 FC Array with sccli 1.6.2 or later
- Sun StorEdge 3511 FC Array (SATA) with sccli 1.6.2 or later
Modification HistoryDate: 12-OCT-2005
12-Oct-2005:
- Updated Contributing Factors and Resolution sections, re-release as Resolved
AttachmentsThis solution has no attachment