Previous | Contents |
This chapter describes changes to the installation process for ACMS
Version 4.3.
2.1 General Installation Information
This kit requires that ACMS be shut down either before or automatically during the installation. (For installation in an OpenVMS Cluster, Compaq recommends that ACMS be stopped on all cluster nodes prior to installation.)
Install this kit with the VMSINSTAL utility by logging in to the SYSTEM account and typing a command in the following format at the DCL prompt, specifying the name of the kit ([kit_name]) and the location of the kit saveset ([saveset_location]):
$ @SYS$UPDATE:VMSINSTAL [kit_name] [saveset_location] |
Table 2-1 and Table 2-2 list the ACMS Version 4.3 kit names:
Kit | Name |
---|---|
Development | ACMSDEVA_043 |
Run-time | ACMSRTOA_043 |
Remote-access | ACMSREMA_043 |
Kit | Name |
---|---|
Development | ACMSDEVV_043 |
Run-time | ACMSRTOV_043 |
Remote-access | ACMSREMV_043 |
After the installation completes, run SYS$STARTUP:ACMSTART.COM on all nodes prior to restarting ACMS.
This installation does not require a reboot.
2.2 CDD Version 6.x Support Restricted
Oracle CDD/Repository Version 6.1 is not supported by ACMS V4.x. An access violation occurs whenever ACMS attempts to access the dictionary.
Oracle CDD/Repository Version 6.1A is supported by ACMS V4.x.
2.3 ACMSTART.COM Installs DECC$SHR and ACMS$TRACE_SHR.EXE and Optionally Executes ACMS$MGMT_SETUP.COM
ACMSHR.EXE now links with the shared image ACMS$TRACE_SHR.EXE, which links with DECC$SHR.EXE. ACMSTART.COM has been modified to install both DECC$SHR.EXE and ACMS$TRACE_SHR.EXE. It is necessary to leave both of these images installed.
If the DECC$SHR image is not installed on the system, you get the following error when starting ACMS:
$ ACMS/START SYS %DCL-W-ACTIMAGE, error activating image DECC$SHR -CLI-E-IMGNAME, image file $100$DUA253:[SYS4.SYSCOMMON.][SYSLIB]DECC$SHR.EXE;1 -SYSTEM-F-PRIVINSTALL, shareable images must be installed to run privileged image |
If ACMS$TRACE_SHR.EXE is not installed on the system, you get the following error when starting ACMS:
$ ACMS/START SYS %DCL-W-ACTIMAGE, error activating image ACMS$TRACE_SHR.EXE -CLI-E-IMGNAME, image file $100$DUA253:[SYS4.SYSCOMMON.][SYSLIB]ACMS$TRACE_SHR.EXE -SYSTEM-F-PRIVINSTALL, shareable images must be installed to run privileged image |
If the ACMS Remote Manager was configured (using
SYS$STARTUP:ACMS_POST_INSTALL.COM), ACMSTART.COM will execute the
procedure ACMS$MGMT_SETUP.COM. This procedure prepares the OpenVMS
environment to run the ACMS Remote Manager. ACMS$MGMT_SETUP.COM is
executed just before ACMS_SETUP.COM.
2.4 ACMS Version Identification File (ACMS_ECO_LEVEL.DAT)
Starting with the ACMS Version 4.0 ECO Kit 1 release, there is a file distributed with the kit called ACMS_ECO_LEVEL.DAT. This file is placed in the SYS$SYSTEM directory during the installation.
The file ACMS_ECO_LEVEL.DAT contains the version identifier for the
ACMS release that is currently installed on the system. This file is
distributed for all releases including full releases, which have an ECO
level of 0. When reporting any problems with ACMS, refer to this file
to determine the current ACMS version that is being used. The version
in this file may be more current than the version numbers that appear
throughout the ACMS system, especially when running with an ECO release.
2.5 Reinstall ACMS After OpenVMS Upgrades
When upgrading major versions of the OpenVMS Alpha operating system, it is necessary to reinstall ACMS. This applies when upgrading the OpenVMS Alpha operating system from Version 6.x to 7.x.
It is not necessary to reinstall ACMS on OpenVMS VAX systems that have
been upgraded from OpenVMS VAX Version 6.x to 7.x.
2.6 Programming Examples on ACMS Kit
The Remote Manager programming examples shipped with the ACMS kit require that the file SYS$COMMON:[SYSHLP.EXAMPLES.ACMSMGMT]ACMSMGMT.OLB be copied to SYS$LIBRARY.
To compile and link the Remote Manager programming examples, execute the following command procedure:
$ @SYS$COMMON:[SYSHLP.EXAMPLES.ACMSMGMT]ACMS$MGMT_EXAMPLES_BUILD.COM |
In the file SYS$COMMON:[SYSHLP.EXAMPLES.ACMSMGMT]ACMS$MGMT_EXAMPLES.C, remove the following sentence from the section called How to build this program:
In addition, you must have access to object module acms$mgmt_get_creds.obj, which is installed in the ACMS$MGMT_EXAMPLES directory.
In the file SYS$COMMON:[SYSHLP.EXAMPLES.ACMSMGMT]ACMS$MGMT_EXAMPLES.C, item 4, change the following LINK command comment:
**** $ LINK ACMS$MGMT_EXAMPLES,ACMS$MGMT_GET_CREDS,ACMSMGMT_RPC_CLNT, ACMSMGMT_RPC_XDR,SYS$INPUT/OPT **** SYS$SHARE:UCX$RPCXDR_SHR/SHARE **** SYS$SHARE:DECC$SHR/SHARE |
to:
**** $ LINK ACMS$MGMT_EXAMPLES,ACMSMGMT_RPC_CLNT,ACMSMGMT_RPC_XDR,SYS$INPUT/OPT **** SYS$LIBRARY:ACMSMGMT.OLB/LIB **** SYS$SHARE:UCX$RPCXDR_SHR/SHARE **** SYS$SHARE:DECC$SHR/SHARE |
The Compaq ACMS Version 4.3 for OpenVMS installation procedure incorrectly calculates the minimum disk space required to install the ACMS Development Kit.
Before attempting to install the ACMS Development Kit, ensure that your system has at least 44,000 blocks of disk space available. If you have insufficient disk space, the installation will fail.
This chapter describes the software problems that have been corrected
in ACMS Version 4.3.
3.1 Unlimited ENQLM Value Allowed
An unlimited ENQLM value is now allowed on OpenVMS Version 7.1 VAX and
Alpha. Setting ENQLM to its maximum value of 32767 in AUTHORIZE
automatically sets ENQLM to the new maximum of 16776959 allowed in
OpenVMS Version 7.x.
3.2 Determination of ENQLM Value
On OpenVMS VAX Version 7.1, the value of ENQLM is now set from the
higher of either the PQL limit or the AUTHORIZE limit.
3.3 Priority of ACMS EXC and Server Processes Now Set Correctly
The base priority of ACMS EXC and server processes is now taken from
the application user's SYSUAF.DAT authorization record.
3.4 Shutdown Delay Corrected
On OpenVMS Alpha Verison 7.1, a 10- to 20-second delay when shutting
down the ACMS ACC process or exiting from an agent has been corrected.
3.5 ADU LINK Command Access Violation (ACCVIO) Corrected
A potential problem that may cause an access violation (ACCVIO) when
linking very large applications with the ADU LINK command has been
corrected.
3.6 Infinite Process Loop Corrected
A potential infinite process loop resulting from an access violation
(ACCVIO) or a reserved operand fault (ROPRAND) for processes linked
with ACMSHR has been corrected.
3.7 ACMS_SWL Image Replaced
On OpenVMS Alpha systems, the ACMS_SWL process sometimes aborted with
an ACCVIO when a large number of errors occurred at the same time. The
ACMS_SWL image, file SWLPROC.EXE, has been replaced to improve
operation.
3.8 ACMS_SWL No Longer Aborts When it Runs Out of Disk Space
Prior to ACMS Version 4.3, ACMS_SWL aborted if the disk on which the SWL.LOG file resides ran out of disk space. In ACMS Version 4.3, ACMS_SWL stops writing SWL entries when it runs out of space. To restart the writing of the error entries, free up the space on the disk where SWL.LOG currently resides, or define the logical ACMS$SWL_LOG (DEFINE/SYSTEM/EXEC) to another disk. Use the command MCR SWLUP RENEW to redirect the output to the new location.
This chapter lists the known problems in ACMS Version 4.3.
4.1 Application May Abort with %SYSTEM-F-BADPARAM Error
The startup of an application may abort with a %SYSTEM-F-BADPARAM error. This could be a result of a high MAXPROCESSCNT parameter value. The ACMS EXC is attempting to create a termination mailbox that will hold a message from every potential process on the system. The formula that ACMS EXC uses is 84 * MAXPROCESSCNT, with an upper limit of 60000 bytes.
The workaround is to reduce the MAXPROCESSCNT parameter to a value less
than 714.
4.2 Incorrect Queue Order with RMS Journaling
When RMS journaling is enabled on an ACMS queue file, elements may be dequeued out of order of insertion. This is due to the way the ACMS$DEQUEUE_TASK service uses the RMS RRL (Ignore Read Lock, RAB$V_RRL bit in the RAB$L_ROP field) to find the next queue element. The queuing service will always attempt to return a queue element or a QUE EMPTY error.
When there are multiple threads in a queue file, there is no guarantee
of the order in which queue elements will be dequeued.
4.3 ADU LINK Aborts with %STR-F-INSVIRMEM or %SYSTEM-F-ACCVIO Error
Linking with a large number of tasks may cause the ADU LINK command to abort with one of the following error messages:
%STR-F-INSVIRMEM, insufficient virtual memory |
or
%SYSTEM-F-ACCVIO, access violation |
This error may be the result of the PGFLQUO parameter being set too
low. To fix the problem, increase the size of the PGFLQUO parameter.
4.4 ACMS SWL Process May be Deleted
Under certain circumstances, the ACMS_SWL process causes an access violation and the process is deleted. This causes ACMS processes that are trying to write information to the Software Event Logger (SWL) log to hang in RWMBX state.
A workaround for this problem is to run the following command procedure as a batch job. This command procedure is not available on your system; you will need to copy it from this document. This batch file monitors the ACMS_SWL every minute (timing can be changed), and restarts the process if it is missing. The batch job should be queued to run under the SYSTEM account (from which SWLPROC is first run).
$ !*************************************************************** $ begin: $ idx = "" $ $ $ loop: $ $ new_id = f$pid(idx) $ if new_id .eqs. "" then goto restart_swl $ full_name = f$getjpi(new_id,"PRCNAM") $ if full_name .eqs. "ACMS_SWL" $ then $ write sys$output "SWL is running, PID is ''new_id'" $ wait 00:01:00.00 ! wait 1 minute $ goto begin $ endif $ goto loop $ $ restart_swl: $ write sys$output "Restarting SWL" $ gosub start_swl $ exit $!*************************************************************** $START_SWL: $ RUN sys$system:SWLPROC - /PRIV=(SYSNAM,SYSPRV,WORLD,OPER,GRPNAM,GROUP)- /PROC=ACMS_SWL- /QUEUE=20- /UIC=[1,4]- /PAGE_FILE=15360- /PRIORITY=8- /DUMP - /OUTPUT=SYS$MANAGER:ACMSSWL.LOG $ RETURN $!*************************************************************** |
If the BYTLM quota is set too low, the ACMS EXC process can disappear without recording an error in either the SWL or Audit Trail Report (ATR) log file and without producing a process dump file. The SWL log shows the following error message, from the ACC process, indicating that the EXC process terminated:
%ACMSACC-E-APPLTERM, EXC process for application <application-name> terminated -Unknown message - error code 00000000 |
However, the EXC process does not log an error to the SWL log to provide additional details of the failure.
This behavior occurs when the system runs out of BYTLM quota. To correct this problem, raise the system's BYTLM quota and reboot the system.
To find a suitable higher value for the BYTLM quota, try doubling the current value. If there is still a problem when using the resulting value, monitor the system as you try different lower BYTLM values until you can find a suitable value.
Monitoring the BYTLM value may not be a straightforward process. The DCL command ANALYZE/SYSTEM can report a BYTLM value that is lower than the value set during SYSGEN. Processes can appear to have a lower allocation of buffered I/O pool (BYTLM) than originally defined for the process, and the allocation can appear to vary.
An article in the remedial kit describes how BYTLM is set and reported by OpenVMS. For more details about the BYTLM quota and the accounting of BYTLM, refer to the article, identified in the remedial kit as follows:
Copyright (c) Digital Equipment Corporation 1996. All rights reserved. PRODUCT: OpenVMS VAX, All Versions OpenVMS Alpha, All Versions COMPONENT: Memory Management SOURCE: Digital Equipment Corporation QUESTION: Why do processes appear to have a lower allocation of buffered I/O pool, BYTLM, than originally defined for the process and why does the allocation seem to vary? |
If the minimum number of servers for an application is nonzero and a STOP/ID is done on an ACMS server process, the Remote Manager incorrectly reports the number of free servers. The free server count is not decremented for the server process being stopped.
The number of free servers will be one more than there actually is.
4.7 Hard Coded Restriction of 4095 Application Starts During an ACMS Invocation
ACMS allows only 4095 application starts during an ACMS invocation.
This restriction is hard coded. There is a check in the ACMS Central
Controller (ACC) (ACCPACKAGES.LIS) for the package ID to be less than
or the same as 4095. This is also used as part of the process name for
the Application Execution Controller (EXC). The name is
ACMS01EXCxxx000, where xxx is between 0 and FFF
(hexadecimal).
4.8 ACMS Task Debugger May Display Incorrect Information on OpenVMS Version 7.x
On OpenVMS VAX and OpenVMS Alpha Version 7.1, the debugger was changed so that it cached the data. When an EXAMINE command was issued, the cached data was displayed.
In the ACMS environment, there are multiple subprocesses. Each
subprocess has its own copy of the debugger, and each copy has the
ability to deposit or examine memory. Because of this, the debugger in
one subprocess does not know that the memory has been changed by
another subprocess and, therefore, displays the incorrect data from
cache.
4.9 ACMS May Fail to Start with %SYSTEM-F-NOSUCHNODE Error
If ACMS is used in a mixed-network environment (Compaq DECnet Phase IV for OpenVMS and Compaq DECnet-Plus for OpenVMS), an ACMS Phase IV DECdtm transaction will fail with a '%SYSTEM-F-NOSUCHNODE, remote node is unknown' error message when it attempts to add the transaction from the DECnet-Plus system.
This problem is a result of the translation of the logical name SYS$DECDTM_NODE_NAME on the DECnet-Plus system. ACMS translates this logical name and returns it to the Phase IV system, which is unable to use the long name.
The workaround is to redefine this logical name to its Phase IV
equivalent.
4.10 Disabling Id and Config Classes Using CLASS=*
It is possible to add a collection record with class of * and a collection state of DISABLED. If the record includes either an entity type or entity name, it overrides the defaults for Id and Config class for that entity.
The applicable processes always populate the Id and Config classes when they start, but subsequent changes to the Config class are not updated.
The following example demonstrates this behavior:
SPARKS> ACMSMGR SHOW COLL ACMS Remote Management -- Command line utility ACMS V4.3 Entity/Collection Table Display Time: 29-SEP-1999 11:51:34.04 Node Wt Entity Collect Collect Type Entity Name Class State --------------------------------------------------------------------- SPARKS 2 * id enabled SPARKS 2 * config enabled SPARKS> ACMSMGR SHOW PROC ACMS Remote Management -- Command line utility ACMS v4.3 Process Table Display Time: 29-SEP-1999 11:51:38.27 Server Entity Collection States Process Name -or- Application.[server_name, Node Type PID task_group_name] Id Cfg RT POOL ------------ -------- -------- ---------------------------------------------- SPARKS acc 202037E0 ACMS01ACC001000 1 1 0 0 SPARKS tsc 202037E2 ACMS01TSC001000 1 1 0 0 SPARKS qti 20203566 ACMS01QTI001000 1 1 0 0 SPARKS cp 20203663 ACMS01CP001000 1 1 0 0 SPARKS cp 20203564 ACMS01CP002000 1 1 0 0 SPARKS cp 20203565 ACMS01CP003000 1 1 0 0 SPARKS> ACMSMGR ADD COLL/ENT=EXC/CLASS=*/COLL_S=DISABLED ACMS Remote Management -- Command line utility Call to add collection on server SPARKS was executed %ACMSMGMT-S-SUCCESS, Operation completed SPARKS> ACMSMGR SHOW COLL ACMS Remote Management -- Command line utility ACMS V4.3 Entity/Collection Table Display Time: 29-SEP-1999 11:52:27.70 Node Wt Entity Collect Collect Type Entity Name Class State ------------ -- -------------------------------------------------------- SPARKS 2 * * id enabled SPARKS 2 * * config enabled SPARKS 3 exc * * disabled SPARKS> ACMSMGR START EXC/APPL=TAPP ACMS Remote Management -- Command line utility Call to start ACMS application tapp on server SPARKS was executed %ACMSMGMT-S-SUCCESS, Operation completed SPARKS> ACMSMGR SHOW PROC ACMS Remote Management -- Command line utility ACMS V4.3 Process Table Display Time: 29-SEP-1999 11:54:40.82 Server Entity Collection States Process Name -or- Application.[server_name, Node Type PID task_group_name] Id Cfg RT POOL ------------ -------- -------- ------------------------------------------- SPARKS acc 202037E0 ACMS01ACC001000 1 1 0 0 SPARKS tsc 202037E2 ACMS01TSC001000 1 1 0 0 SPARKS qti 20203566 ACMS01QTI001000 1 1 0 0 SPARKS cp 20203663 ACMS01CP001000 1 1 0 0 SPARKS cp 20203564 ACMS01CP002000 1 1 0 0 SPARKS cp 20203565 ACMS01CP003000 1 1 0 0 SPARKS exc 20201668 ACMS01EXC002000 0 0 0 0 SPARKS server TAPP.TSERVER 1 1 0 0 SPARKS group TAPP.T_GROUP 1 1 0 0 |
ACMSCFG causes an access violation with the following syntax:
NODE> ACMSCFG SET PARAM/RPC_STACKSIZE = 10 |
If the spaces are removed (/rpc_stack_size=10), the command works.
This problem exists for all parameters specified in this command only. The following is an example of the access violation.
%ACMSMGMT-E-BAD_VALUE, Invalid qualifier, cannot interpret value rpc_stacksize %SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=00000000, PC=000DDE0C, PSL=0BC00000 %TRACE-F-TRACEBACK, symbolic stack dump follows module name routine name line rel PC abs PC 000DDE0C 000DDE0C MGMTCONFIG_CMD parse_args 19710 000000D5 0000A1DD MGMTCONFIG_CMD update_param_rec 21903 0000007A 0000F2A6 MGMTCONFIG_CMD main 18640 000008E4 000079E4 |
Previous | Next | Contents |