Problems with sapstartsrv as of 7.00 & 6.40 patch 169

As of 7.00, the SAP start service "sapstartsrv" is delivered on all platforms. The previous DCOM "SAPControl" interface was replaced with a new Web service interface featuring enhanced functions for J2EE Engine administration.

Known issues with the new sapstartsrv start service:

      1. In the case of a double stack (ABAP+Java) instance, the MMC has the status "green" in the "process list" for all processes even though the J2EE Engine does not fully run (yet).

      2. On some Linux distributions, sapstartsrv crashes immediately after the start, or it restarts an incorrect environment for the SAP instance, or its Web service interface hangs and does not respond to requests (for example, from the MMC).

      3. On Solaris, sapstartsrv hangs when it is started as a daemon (" -D" Option) as soon as it tries to determine the environment again (for example, before starting an instance).

      4. A handle leak in sapstartsrv results in accumulating open thread handles, in particular, when closing the connection to the client (SAP MC, sapcontrol, SAP MMC).

      5. In the UNIX syslog, you find sapstartsrv entries with the message "Unable to open trace file sapstartsrv.log" or "Web service interface failed, sapstatrsrv stopped". Nevertheless the service is started and works correctly.

      6. The "trusted connect" of the sapcontrol for the sapstartsrv (Option " -prot NI_HTTP") fails on Unix with the error "FAIL: HTTP error, HTTP/1.1 401 Unauthorized", even though the user adm was used.

      7. On Windows, the self-restart service does not work (for example, after a sapstartsrv.exe update, or using the Web service interface). The service closes properly, but is not restarted.

      8. The system fails to authenticate a Windows user, even though the user and password have been correctly specified.

      9. The job object specific to the SAP instance created by the Windows sapstartsrv is not displayed in Windows performance manager (perfmon), for example.

      10. Sapstartsrv may terminate when reading the ABAP syslog. Therefore, MMC and other clients cannot display the syslog.

      11. Sapstartsrv may terminate when the J2EE engine is in "safe mode".

      12. The automatic service restart and the manual service restart using the WebService interface (for example when you use sapcontrol) do not work on Windows. Instead, the service simply terminates.

      13. Sapstartsrv periodically returns "unaligned access to" messages in /var/log/messages on 64-bit Linux systems.

      14. Improvements to the security of the Web service interface of sapstartsrv.

      15. sapstartsrv will not start on Unix systems, even though you have entered it correctly in /usr/sap/sapservices or it has been started correctly by startsap. sapstartsrv may start, but loads the required shard libraries from the central /usr/sap//SYS/exe/run executable directory.

      16. Even though LDAP registration is active (ldap/autoregister=1), sapstarstrv does not register itself in the LDAP Directory and no dev_ldaps log file is written.

      17. After running on UNIX systems for some time, sapstartsrv no longer replies to requests from new clients (MMC, sapcontrol).sapcontrol returns the error "FAIL: End of file or no input: 'Operation interrupted or timed out'".

      18. On Windows, sapstartsrv inadvertently recognizes that its executables are being updated and restarts itself unnecessarily, which may cause follow-up problems, especially with HA installations.

      19. When reading Java logfiles using the WebService interface of the sapstartsrv, the system cannot find all entries when you use filters (for example, in the logviewer of the NWA).

      20. The J2EE Engine writes log file entries in the new 2.0 format. These cannot be read by clients (SAP MMC, SAP MC, sapcontrol, NWA) using the Web service interface of sapstartsrv.

      21. The response times of the NWA log viewer when reading J2EE log files using the sapstarstrv Web service interface are still poor even though a restrictive time filter was set.

      22. When you use the sapstartsrv Web service interface (for example, from SAP MMC/MC) for the log file analysis, time stamps from SAP developer traces may not be evaluated correctly, which means that some log file entries may appear with the incorrect times in the result.

      23. When you use the sapstartsrv Web service interface (for example, from SAP MMC/MC or NWA) when reading J2EE log files, specific log file entries are not found, and sapstartsrv may terminate.

      24. If very long profile parameter values (more than 1024 characters) have been set, sapstartsrv does not work correctly or terminates. Ususally, long profile parameter values are generated from long environment variables (such as "PATH"), due to entries in the profile such as "SETENV_00 = PATH=$(DIR_EXECUTABLE);%(PATH)".

      25. On Unix in 6.40, 7.00 and 7.01, a command that was started using the OSExecute method of the WebService interface does not recognize the finishing of child processes started by it, and waits until timeout is reached.

      26. On UNIX, stopped processes may be reported as running if another process with the same process ID (PID) is started at a later stage.

      27. On UNIX, a "trusted connect" (see 927637) is no longer possible when the service/host name is set.

      28. On Unix (observed on Solaris), sapstartsrv hangs, the memory requirement increases rapidly and the CPU load is high. This occurs if a webservice client (for example, NWA) tries to authenticate with an empty user or password.

      29. In a Windows HA cluster, the instance or system "restart" function, for example, in SAP MMC, does not work correctly. Instead, a HA failover occurs. A stop followed by a start however works correctly, without triggering a HA failover.

      30. On Windows, the processes of the instance may be started with an incomplete environment of sapstarstrv if the Web service method "GestInstanceProperties" is called during the start process. For example, this is quite likely if the start was carried out using the "StartWait" function of sapcontrol. The consequences may vary; for example, the setup of the database connection of the ABAP work processes may fail if environment variables are missing.

      31. If an incorrect SSL client connects to the sapstartsrv https port, no further SSL clients may be edited until the incorrect client closes its network connection.

      32. On Unix, the sapstarstrv hangs when the environment is restarted or it restarts the environment incorrectly if the shell prompt of the adm user contains "=" or "env" (for example, set using "prompt" or "PS1").

      33. On some Unix platforms (observed on HPUX and Solaris), due to a deadlock, sapstartsrv hangs (when the environment is ended or restarted) and is not closed.

      34. The Web Services Description Language (WSDL) of the Web service interface statically includes "http://localhost:50013" as the address location. At runtime, it is not dynamically adjusted to the actual log, host name, and port.

      35. sapstartsrv reports partially incorrect data for AS ABAP, AS Java or double stack instances, for example, if a yellow status is reported for the ABAP dispatcher (disp+work) or jcontrol or jstart with descriptions such as "Dialog Queue info unavailable" or "J2EE status info unavailable". Other information such as "ABAP WP Table", "Queue Statistic" or "Syslog" may also be affected.

      36. When calling the AnalyseLogfiles Web service method, sapstartsrv may crash if J2EE log files exist that do not belong to a file set.

      37. During the start, sapstartsrv occasionally crashes if the CCMS agent is active (as of 7.10).

      38. If an instance contains many log files, the speed when listing and reading the log files may be reduced considerably. In some clients (for example, NWA), this may result in a timeout being exceeded so that log files can no longer be read.

      39. On Windows, a 701 sapstartsrv does not respond; clients such as SAP MMC / MC or sapcontrol hang. "netstat -ano | findstr 5..13 |findstr LISTENING" shows two processes that respond to the same port. For one of them, a PID of a sapstartsrv that no longer runs is displayed.

      40. The J2EEControlCluster Web service method ignores dual-stack instances with an integrated message server and enqueue. For example, if the Java deployment uses this to restart the AS Java cluster, the relevant instance is not restarted or the error message "No matching J2EE instance found" occurs if no further AS Java instances exist in the system.

      41. In the host agent mode of sapstartsrv, the Web service method ReadLogFile delivers "Invalid filename" in 720 patch levels 28 to XX, even though the specified file name of the log file is correct and is delivered correctly when "ListLogFiles" is called, for example. This affects log files that were specified using directory names by means of the profile parameter service/logfile_XXX.

      42. When it reads back J2EE log files, ReadLogFile returns an incorrect result (too many log entries) on reaching the start of the file. Therefore, Web service clients (such as the NWA Log Viewer) may display an incorrect result when scrolling back in a log file.

      43. As of 8.00, the ReadLogFile method also supports the generic log file format (*.glf files) that is used by Business Objects products, for example. Older sapstarstrv are not able to parse these files and return the content as unstructured text as a result.

      44. On Windows, child process that are started temporarily by sapstartsrv (for example, sapstartsrv with a "-sapparam" command lines parameter) remain, even though the sapstartsrv service has been stopped.

      45. On Windows, the method ReadLogFile returns incorrect results when reading Unix-style text files backwards.

Other terms

sapstartsrv, MMC, Microsoft Management Console, sapcontrol, Web service, NWA, SAP NetWeaver Administrator
Solution

      1. As of 7.00 patch level 29, sapstartsrv also checks the status of the J2EE Engine for double-stack instances and assigns it to the dispatcher node in the "Process list".

      2. As of 7.00 patch level 30, sapstartsrv uses "openpty" on Linux instead of opening the devices directly, since this is no longer possible on some Linux distributions.

      3. This problem is solved as of 7.00 patch level 38.

      4. This problem is solved as of 7.00 patch level 38.

      5. The Unix startsap tries to restart sapstartsrv even though it was started already using the entries in the /usr/sap/sapservices script. However, sapstartsrv recognizes this and closes itself immediately and writes a descriptive message to the UNIX syslog. You can ignore the message because it does not refer to an error.

      6. The Unix sapstartsrv unintentionally deletes its Unix domain socket (/tmp/.sapstream513) for example, when stopping the instance. The "trusted connect" using the Unix domain socket no longer works. The error is corrected in the sapstartsrv as of 7.00 patch level 48.

              Even though the Unix 6. 40 sapstartsrv creates the Unix domain socket (/tmp/.sapstream513) correctly, it does not recognize this itself. Therefore, the sapstartsrv.log does not contain the line:

              Trusted http connect via Unix domain socket '/tmp/.sapstream5XX13' enabled.

              and the trusted connect does not work. This error is corrected in sapstartsrv as of 6.40 patch level 204.

      7. As of 7.00, the Windows service user SAPService no longer has administrator authorization. Sapstartsrv does not prompt for required authorizations when the service restarts. The error is corrected in sapstartsrv as of 7.00 patch level 55.

      8. The Windows sapstartsrv executes a user authentication using LOGON32_LOGON_NETWORK.  The user being authenticated may not have the required authorization (for example, if the setting "Deny access to this computer from the network" was set explicitly for the Windows user). As of 7.00 patch level 58, sapstartsrv also attempts an authentication using LOGON32_LOGON_INTERACTIVE, if necessary.

      9. As of 7.00, the SAPService user is no longer a member of the local administrator group, and therefore must explicitly set access authorizations for the job object that was created. Use 7.00 patch level 58 or higher.

      10. If the syslog contains incorrect entries (message data do not match the format definition (se92)), sapstartsrv may terminate when it attempts to format them. As of 7.00 patch level 60, the implementation of the message formatting in sapstartsrv is more robust.

      11. The error is corrected as of 7.00 patch level 75.

      12. As of 7.00, the SAPService user is no longer a member of the local administrator group. This means that the user normally lacks the authorization required to start or stop a Windows service. As of 7.00 patch level 75, when you install the service and sapstartsrv registers itself, sapstartsrv sets the ACL for the service so that the SAPService user is allowed to start and stop the service. Use 7.00 patch level 75, and register the service again.

      13. The error is corrected as of 7.00 patch level 75.

      14. We have added corrections that improve the security of the Web service interface as of 7.00 patch level 75.

      15. When you start sapstartsrv using the "-D" option (as usual) as a daemon, the user environment is reset and forked into a new sapstartsrv process. Only the user environment is reset. Any previously set enhancements of the shared library search paths (LD_LIBRARY_PATH, LIBPATH, SHLIB_PATH) are lost. Therefore, shared libraries may not be loaded, or may be loaded incorrectly from the central /usr/sap//SYS/exe/run executable directory instead of from the instance-local executable directory. As of 7.00 patch level 85, sapstartsrv inserts its own executable directory in the shared library search path, in order to ensure that all shared libraries can be loaded locally.

      16. 6.40 patch level 168 and 7.00 patch level 96 damaged the LDAP registration on Windows. This error is corrected again as of 6.40 patch level 175 and 7.00 patch level 102.

      17. As of 6.40 patch level 190 and 7.00 patch level 118, a resource leak has been corrected in sapstartsrv. This resource leak caused the problem described above on some platforms (in particular HPUX).

      18. This problem is solved with 6.40 patch level 199 and 7.00 patch level 126. Before a self-restart, the system compares file versions in addition to the time stamps of the files sapstartsrv.exe and sapstartsrv.exe.new.

      19. The error is corrected with 640 patch level 211, 700 patch level 138 or 710 patch level 82.

      20. With s 701, 710 patch level 96, and 711, sapstartsrv supports the 2.0 J2EE log file format.

      21. As of s 701, 710 patch level 96, and 711, sapstartsrv has an optimized J2EE log file access if a time interval is set for filtering.

      22. The error is corrected with 701 Patch 9, 710 Patch 115, 711 Patch 3.

      23. The error is corrected with 701 Patch 9, 710 Patch 115, 711 Patch 3.

      24. The error has been solved with 640 patch level 252, 700 patch level 179, 701 patch level 11, 710 patch level 122 and 711 patch level 14.

      25. The error in sapstartsrv is corrected with 640 patch level 252, 700 patch level 181 or 701 patch level 14.

      26. The error in sapstartsrv is corrected with 701 patch level 17, 710 patch level 128 or 711 patch level 17.

      27. The error in sapstartsrv is corrected with 701 patch level 17, 710 patch level 128 or 711 patch level 17.

      28. Depending on the Unix PAM configuration, an empty user name or password may not be accepted, and an endless loop with memory leak occurs. The error is corrected with 701 patch level 30, 710 patch level 139, 711 patch level 26.

      29. The error is corrected with 701 patch level 30, 710 patch level 139 and 711 patch level 26. As of 710, you must also update sapNThalib.dll.

      30. The error is corrected with 701 patch level 33, 710 patch level 142 and 711 patch level 29.

      31. The error is corrected with 701 patch level 39, 710 patch level 148 and 711 patch level 35.

      32. The error is corrected with 640 patch level 286, 700 patch level 210, 701 patch level 48, 710 patch level 157, 711 patch level 44 and 720 patch level 5.

      33. If you use 701 patch level 51, 710 patch level 160, 711 patch level 47, or 720 patch level 8, the deadlock in the (CCMS) atexit handler is prevented if a fork runs at the same time.

      34. If you use 701 patch level 51, 710 patch level 160, 711 patch level 47 or 720 patch level 8, sapstartsrv creates a dynamic address location based on the host header field of the HTTP request of the client.

      35. Some monitoring data is determined by sapstarstrv from the shared memory of the instance. If processes of sapstartsrv and the instance use different parameterization, the functions can be negatively affected. As of 710, the separate start profile was removed, so that only a joint parameterization is used, which in turn avoids the problem. As of 640 patch level 295, 700 patch level 219, 701 patch level 30 and 720 (important for 702), sapstartsrv recognizes that it was started with a start profile and additionally loads the instance profile, which also avoids the problem.

      36. The error is corrected as of 701 patch level 71, 710 patch level 179, 711 patch level 65, and 720 patch level 27.

      37. The error is corrected as of 710 patch level 179, 711 patch level 65, and 720 patch level 27.

      38. We generally recommend that you keep the number of log files manageable, delete unnecessary files (if applicable), and limit the archiving of AS Java log files to a sensible amount. As of 720 patch level 28, the performance of the Web service methods ListLogFiles and ReadLogFile was also further improved.

      39. On Windows, the 701 sapstartsrv inadvertently passed on its network ports to the lower-level processes. When you restart sapstarstrv when the instance is running (for example, automatically in the case of a kernel patch or manually using SAP MMC), the port remains blocked until all lower-level processes (=instance processes) have been completed. The error is corrected as of 701 patch level 73. When you apply the kernel patch, the error occurs one last time on each SAP instance the next time the instances are started after sapstartsrv has restarted. To cleanup this state, you must subsequently stop and then restart all of the processes of each affected SAP instance. Since you cannot currently use SAP Microsoft Management Console (MMC) due to this error, the easiest thing to do is to stop the affected SAP Windows services within the Microsoft Computer Management MMC SnapIn. IMPORTANT: This stops the affected SAP instance. You must subsequently restart the SAP services.

      40. The error is corrected as of 720 patch level 29.

      41. The error is corrected as of 720 patch level 35. Update your host agent accordingly.

      42. The error is corrected with 640 patch level 320, 700 patch level 245, 701 patch level 82, 710 patch level 189, 711 patch level 76, and 720 patch level 35.

      43. As of 720 patch level 37, the generic log file parser is also integrated into the 720 sapstartsrv. Update your host agent accordingly.

      44. The error is corrected as of 720 patch level 45.

      45. The error is corrected as of 720 patch level 46.