Loading...
Loading...
OUTAGE PREVENTION NOTIFICATIONThis notice describes conditions that may result in an application outage and actions you can take to prevent the outage. Study this information carefully to evaluate how much risk, if any, this behavior could present in your environment. Consider taking the preventive action described.SUMMARY:A recurring %7440 processor halt will likely occur if a backup process is created when enough processes have already been created since coldload to cause process sequence numbers (also referred to as process verifiers) to reach a value greater than or equal to 2^32.Note that the system is not at risk for this problem immediately after a coldload but data collected from HPE customers suggests that the risk of the problem described in this OPN can arise after a period as short as a month or two after coldload with the affected SPRs listed below. Therefore, HPE encourages customers running these SPRs to determine if they may be at risk based on a calculation method described later in this OPN.The affected SPRs are the T9050L02^BBN and T9050L02^BBQ SPRs, which are L25.02-based. Although this defect is present in the L25.02.00 and L25.02.01 RVUs, T9050L02^BBN and T9050L02^BBQ are usable on RVUs from L20.05 to L25.02. If you have installed either of these SPRs on an older RVU, this issue applies to your system also. The problem can be avoided by applying T9050L02^BBR or a later SPR.DETAILED DESCRIPTION:In a system running either the T9050L02^BBN or T9050L02^BBQ SPRs, a processor may halt with a %7440 halt code when a newly created primary process queries its backup process handle after the system has been up and running for an extended period of time. The actual amount of time depends mainly on the number of processes that have been created since coldload.The Nonstop OS assigns unique sequence numbers (also referred to as verifiers) to processes at process creation time. The process sequence number or verifier is a 64-bit value used to differentiate a current process instance from previous instances with a matching processor and PIN. This verifier increases over time, and it is stored in an internal Nonstop OS per-process data structure referred to as the process handle. For more details, refer to section Process Handles in the Guardian Procedure Calls Reference Manual.A process-pair process may query the process handle of its brother process. This is a typical operation when a backup process is created, and it entails checking the sequence number of the backup process. In a system running either the T9050L02^BBN or T9050L02^BBQ SPRs, the sequence number verification produces an incorrect outcome if the sequence number has reached a value greater than or equal to 2^32. This results in an inadvertent %7440 processor halt.Attempts to reload the halted processor, followed by attempts to create the backup process on that processor, will trigger further occurrences of the %7440 processor halt, thereby preventing the halted processor from remaining online. Once this situation occurs, a system load is required for recovery.Customers with systems running T9050L02^BBN and T9050L02^BBQ are strongly recommended to upgrade to T9050L02^BBR (available 2nd half of Nov 2025) at an opportune time to prevent this halt from happening in the future.T9050 SPRs other than T9050L02^BBN and T9050L02^BBQ do not have this issue.To find out if a system is at risk:Run a named TACL. For example:TACL /name $abc1/After the logon, find the CPU and PIN number of such TACL process.For example:> status $abc1System \CORFUProcess Pri PFR %WT Userid Program file Hometerm$ABC1 1,85 147 004 255,255 $SYSTEM.SYS16.TACL ...Swap File Name: $SYSTEM.#0 ...Run the following TACL command using the CPU and PIN:$SYSTEM SYS16 12> #convertphandle /INTEGERS/ 1,85#convertphandle /INTEGERS/ 1,85 expanded to:512.63.1.85.0.0.34752.28209.0.43If either the 5th or 6th values are non-zero then you are already at risk for this issue. Otherwise, if both the 5th or 6th values are 0, as in the example above then do the following calculation to estimate the number of days until the system is at risk for this issue:days-since-coldload * (65535/7th-INT16 – 1) = days-until-at-riskWhich in the example above, assuming 66 days since coldload, would be:66 * (65535/34752 – 1) = 58 days until the system is at riskNote that the estimation method above assumes that the daily rate of process creation since the system was cold loaded is consistent. You can use the method described above periodically (for example, once a day or week) to monitor how the days-until-at-risk estimate progresses over time.Likelihood: Likely after a system has been up and running for an extended period.
Affected version(s):T9050L02^BBN and T9050L02^BBQRelease Version Updates (RVUs) containing the defective versions:L25.02Additional RVUs potentially affected because one or more of the SPRs listed above is usable with them:L20.05 - L24.08
RVUs containing the fix for this issue:L25.09L25.02.02 (RVUR available to request in 'Scout for Nonstop Servers')PREVENTIVE ACTION:Workaround: Doing a coldload restarts the process verifiers at lower values, effectively resetting how long until the system would again be at risk until the T9050L02^BBR SPR can be applied to the system.HPE recommends doing the calculation to estimate the number of days until the system is at risk before deciding on doing a coldload for this purpose.Fix:Install theT9050L02^BBRSPR (SPR available to download in 'Scout for Nonstop Servers').This SPR is usable with the following RVUs:L20.05 - L25.02This SPR is present in the following RVU:L25.02.02Note: Refer T9050L02^BBR's softdoc's Requisite SPRs list when installing this SPR on an RVU prior to L25.02Installation impact:SYSGEN and system load required.Please refer to the appropriate softdoc for detailed SPR information, including installation instructions, superseded SPRs, and requisite SPR lists.Appendix:A sample TACL macro is provided at the following URL:https://support.hpe.com/hpesc/public/docDisplay?docId=dp00007064en_usThis macro starts a process, checks its process handle, and does the calculation discussed earlier in this HotStuff:Using the example discussed earlier in this HotStuff, saving the TACL macro as a file named riskchk and running it produces the following output:Note that this script can be executed by anyone who has access to the system.TACL>riskchkChecking risk..Phandle is 512.63.1.85.0.0.34752.28209.0.43Days since coldload: 665th INT16 value:06th INT16 value:07th INT16 value:34752*** Only applicable if the system is running T9050L02^BBN or BBQEstimated days until risk: 58Revision HistoryDocument VersionRelease DateDetails4November 20, 2025Added availability of fix SPR T9050L02^BBR and L25.02.02 RVUR3November 14, 2025Sample TACL Macro moved to a separate file to preserve the code syntax. Provided a link to the Macro in the advisory.2November 07, 2025Revised the availability date of the fix SPR T9050L02^BBR. Added a workaround for consideration until T9050L02^BBR or a later SPR can be applied to the system. Simplified the description on how to estimate the number of days left until a system is at risk of receiving the %7440 halt discussed in this OPN. Provided a sample TACL macro to perform the calculation.1October 31, 2025Initial Version
Operating Systems Affected:NonStop OS L20.05, NonStop OS L20.10, NonStop OS L21.06, NonStop OS L21.11, NonStop OS L22.09, NonStop OS L23.08, NonStop OS L24.08, NonStop OS L25.02
Click on a version to see all relevant bugs
Hewlett Packard Enterprise Integration
Learn more about where this data comes from
BugZero Plan
Streamline upgrades with automated vendor bug scrubs
BugZero Prevent
Wish you caught this bug sooner? Get proactive today.