Asset ID:	1-77-1019604.1
Update Date:	2012-07-18
Keywords:

---

Solution Type  Sun Alert Sure

Solution  1019604.1 :   On Certain Sun Enterprise, Blade, and Netra Servers, TOD may Advance as Much As One Second Every Fifteen Minutes  


PreviouslyPublishedAs
242106

Bug Id
<SUNBUG: 6676309>, <SUNBUG: 6682970>, SUNBUG: 6674205, SUNBUG: 6667409

Product
Sun SPARC Enterprise T5120 Server
Sun SPARC Enterprise T5220 Server
Sun SPARC Enterprise T5140 Server
Sun SPARC Enterprise T5240 Server
Sun Blade T6320 Server Module
Sun Netra CP3260 ATCA Blade Server
Sun Netra T5220 Server

Date of Resolved Release
17-Sep-2008

***Checked for relevance on 18-Jul-2012***

1. Impact

On Sun SPARC Enterprise servers T5120, T5220, T5140 T5240, Sun Blade server T6320, and Sun Netra T5220 and CP3260 Servers, the "Time-of-Day" (TOD) on the system may advance as much as one second every fifteen minutes. Solaris and user applications (ie database applications) which use Solaris TOD may be adversely affected in that event timestamps may be incorrect, potentially affecting data integrity.

2. Contributing Factors

This issue can occur on the following platforms:

SPARC Platform

• Sun SPARC Enterprise T5120 and T5220 Servers with firmware revision prior to 7.1.3.d (without patch 136932-02)
• Sun SPARC Enterprise T5140 and T5240 Server with firmware revision prior to 7.1.3.d (without patch 136936-03)
• Sun Blade T6320 with firmware revision prior to 7.1.0.g (without patch 136933-04)
• Netra T5220 with firmware revision prior to 7.1.3 (without patch 136934-02)
• Netra CP3260 with firmware revision prior to 7.1.3.a (without patch 136935-01)

Note: x86 platform systems are not affected by this issue.

To determine the Firmware revision on the system, log into the Service Processor or CMM as 'root' and run the following:


If a Service Processor /CMM admin account has been created, then log in to that account and run the following:


and


3. Symptoms

Solaris' TOD will advance faster than standard clock as much as one second every fifteen minutes.  This means it can advance 90 seconds in a 24 hour period.  This could cause database entries to have incorrect timestamps, which may affect data integrity.

Due to "spread spectrum" enabled in the hardware clock, the sys-tick frequency VBSC passed on to Solaris does not match the clock frequency, causing an inaccurate calculation of Solaris time resulting in clock drifts of more than the acceptable parameter of less than 2 seconds a day.

4. Workaround

(For Sun SPARC Enterprise T5120 and T5220 Servers only)

To work around the described issue, edit "/etc/system" to include/change the following parameters:

If it is a 1.2GHz system, set:


If it is a 1.4GHz System, set:


then run "init 0" and reboot Solaris.

To determine if the system is 1.2GHz or 1.4GHz you can review SP console logs for the POST report which will list the speed.

Also, prtdiag -v in Solaris will indicate the speed, as in the following:


The Solaris TOD drift will no longer occur after "/etc/system" is edited and 'init 0' is run and the Solaris reboot takes place.

5. Resolution

This issue is addressed on the following platforms:

SPARC Platform

• Sun SPARC Enterprise T5120 and T5220 Servers with firmware 7.1.3.d (as delivered in patch 136932-02 or later)
• Sun SPARC Enterprise T5140 and T5240 Servers with firmware 7.1.3.d (as delivered in patch 136936-03 or later)
• Sun Blade T6320 with firmware revision 7.1.0.g (as delivered in patch 136933-04 or later)
• Netra T5220 with firmware revision 7.1.3 (as delivered in patch 136934-02 or later)
• Netra CP3260 with firmware revision 7.1.3.a (as delivered in patch 136935-01 or later)

Modification History
15-Oct-2008: Clarification to FW version listed in Contributing Factors and Resolution sections
18-Jul-2012: Maintenance update; Checked for relevance

References

SUNPATCH:13693202
SUNPATCH:13693304
SUNPATCH:13693402
SUNPATCH:13693501
SUNPATCH:13693603

Attachments

This solution has no attachment