Sun Fire 3800/4800/4810/6800 and E4900 and E6900 servers do not always maintain their hardware clock correctly which may create issues for the Network Time Protocol (NTP) function and "date -a" command. The potential for data loss involves anything that relies on a consistent time stream, such as a transaction log. For example, replaying a transaction log with a negative time increment in it could potentially lose data if any such application assumes (not unreasonably) that all time increments are positive.

This issue can occur in the following releases:

SPARC Platform

• Solaris 8 without patches 108528-16 and 114526-01
• Solaris 9 without patches 112987-01 and 114526-01

OR systems on the following platforms:

• Sun Fire 3800/4800/4810/6800 with firmware revision prior to 5.19.0 and without patch 114526-01
• Sun Fire E4900 and E6900 Servers with firmware revision prior to 5.19.0 and without patch 114526-01

This issue will only occur when these conditions are present:

1. SNTP is enabled on the System Controller AND
2. NTP is enabled in the domain(s)

To determine the firmware revision of the System Controller (SC), use the following command from the platfom shell of the main SC:

    sc0:SC> showsc
    SC: SSC0
    Main System Controller
    ...
    ScApp version: 5.15.3
    RTOS version: 32

Should the described issue occur, "time step" messages similar to the following may be output to the "/var/adm/messages" log file:

    # egrep ntp /var/adm/messages
    ...
    xntpd[253]: [ID 774427 daemon.notice] time reset (step) -2.059495 s
    xntpd[253]: [ID 774427 daemon.notice] time reset (step) 2.062533 s

along with any of the following messages in the System Controller platform log:

    sc0:SC> showlogs
    ...
    Platform.SC: SNTP bring clock forward by 1 seconds
    Platform.SC: SNTP bring clock backward by 1 seconds
    Platform.SC: System clock has drifted. Resynced with SNTP server

To work around the described issue, isolate the domain from the virtual hardware clock by adding the following to "/etc/system":

    set tod_broken = 1
    set dosynctodr = 0

Note: A reboot may be required in order for this change to take effect.

This issue is resolved in the following releases:

SPARC Platform

• Solaris 8 with patches 108528-16 or later AND 114526-01 or later
• Solaris 9 with patches 112987-01 or later AND 114526-01 or later

AND:

• System Controller Firmware (5.19.0) patch 114526-01 or later

Notes:

1. Both patches for either Solaris 8 or 9 must be installed to resolve this issue.
2. BugID 4966931 was fixed in the 5.18.0 firmware provided by patch 114525-01, however, for complete resolution including the fix to BugID 5089758, it is necessary to install the 5.19.0 firmware provided by 114526-01.

• Updated Contributing Factors for firmware patch and revision clarifications
• Updated Resolution for firmware patch and revision clarifications

• Final patch additions/corrections; re-release as resolved

29-Jul-2005:

• Update Resolution section; reopened, State changed to Workaround

16-Aug-2005:

• Updated Contributing Factors and Resolution sections

This solution has no attachment