| Previous | Contents |
The following problem applies to the ACMS Remote Manager.
4.12.1 Some Collection Table Update Messages May Get Lost
Under some circumstances, dynamic changes to the collection table are not always communicated to all ACMS processes. When this occurs, the ACMS processes that did not receive the update messages will not be synchronized with the collection table. This is most likely to occur for a short period of time (approximately 1 minute or less) whenever the ACMS Trace process (process name ACMS$TRACE_MON) is started.
The Trace process is started under the following conditions:
The workaround is to delay updating the collection table for 30 seconds to 1 minute after the Trace monitor is started.
For example, with the following set of commands, collection messages will be lost if the commands are executed serially with little or no time between commands, as in a command procedure:
$ ACMS/START SYS $ @SYS$STARTUP:ACMS$MGMT_STARTUP $ ! WAIT FOR THE rm TO STARTUP $ WAIT 00:00:05 $ ! $ ACMSMGR ADD COLL/ENT=*/NAME=*/CLASS=RUNTIME/COLL_S=ENABLED |
The problem is uncovered by displaying the process collection states, as in the following example:
$ ACMSMGR SHOW PROCESS
ACMS Remote Management -- Command line utility
ACMS T4.3-1 Process Table Display Time: 24-SEP-1999 14:45:37.75
Server Entity
Node Type PID Process Name -or-
Application.
[server_name, task_group_name] ID Cfg RT POOL
-------- ----- -------- ------------------------------------------------
SPARKS acc 20200D4F ACMS01ACC001000 1 1 0 0
SPARKS tsc 20200D51 ACMS01TSC001000 1 1 0 0
SPARKS cp 20200D52 ACMS01CP001000 1 1 0 0
SPARKS cp 20200D53 ACMS01CP002000 1 1 0 0
SPARKS cp 20200D54 ACMS01CP003000 1 1 0 0
|
In this example, all processes have the run-time class disabled even though a collection table entry was just added to enable them all.
You can take one of two possible actions. First, you can prevent this situation from occurring by delaying slightly before issuing the command to update the collection table. For example:
$ ACMSMGR START SYS $ @SYS$STARTUP:ACMS$MGMT_STARTUP $ ! WAIT FOR THE RM TO STARTUP $ WAIT 00:00:05 $ ! NOW WAIT FOR TRACE TO GET CONNECTED $ WAIT 00:00:40 $ ACMSMGR ADD COLL/ENT=*/NAME=*/CLASS=RUNTIME/COLL_S=ENABLED |
Alternatively, you can correct the situation by applying a change to the collection table. This forces the Remote Manager to evaluate and synchronize the collection states of each process, as in the following example:
$ ACMSMGR SET COLLECTION/ENT=*/NAME=*/CLASS=RUNTIME/COLL_S=ENABLED. |
Information about the LINK/STDL command qualifier was inadvertently
omitted from the ACMS Version 4.3 online help. This command is fully
documented in the Compaq ACMS for OpenVMS ADU Reference Manual.
4.14 Documentation Errors
This section contains errors in the ACMS documentation.
4.14.1 Compaq ACMS Version 4.3 for OpenVMS Installation Guide
The Compaq ACMS Version 4.3 for OpenVMS Installation Guide contains the following errors:
$ SYS$STARTUP:ACMSDI$TARTUP.COM |
$ SYS$STARTUP:ACMSDI$STARTUP.COM |
SYS$LIBRARY:FORMS$MANAGER.EXE |
In Appendix A, Section A.1, Remote Manager Server, of the Compaq ACMS for OpenVMS Remote Systems Management Guide, add the following logical name:
ACMS$MGMT_DISABLED When defined to be T,t,Y,y,1 (or any odd number), this logical prevents ACMS from performing any operations to support Remote Management. Management global sections are not created, and no class data is collected. This logical must be defined /SYSTEM /EXEC. |
This chapter lists the restrictions that apply to ACMS Version 4.3.
5.1 New Restrictions in ACMS Version 4.3
The following restrictions are new in ACMS Version 4.3.
5.1.1 SNMP GETNEXT Calls Return Unpredictable Results
SNMP getnext calls are not supported in this release of the ACMS Remote
Manager. SNMP getnext calls issued to the Remote Manager may appear to
complete successfully; however, the data should not be considered to be
accurate.
5.1.2 Remote Manager is Unusable Following TCP/IP or UCX Shutdown
The ACMS Remote Manager depends on the underlying network software in order to communicate with external processes. If the IP network is stopped, the existing network connections between the ACMS Remote Manager and the IP network become orphaned and cannot be reconnected when the network is restarted. This prevents further communications between external processes and the ACMS Remote Manager.
The workaround is to ensure that the Remote Manager is stopped prior to
stopping the IP network, and to restart it only after the IP network
has been restarted. If the Remote Manager is left running when the
network is stopped, it is acceptable to stop it using the DCL command
STOP/ID.
5.1.3 Error Message When Stopping Remote Manager
When the Remote Manager process is stopped, a detached-process termination accounting record is written for the ACMS$MGMT_SVR process. The final status recorded will be %x00FE9F49, which is translated in the SYS$MESSAGE:ACMSMSG.EXE file as:
%ACMSMGMT-S-SUCCESS, Operation completed. |
This status indicates a successful shutdown of the ACMS Remote Manager.
5.1.4 Remote Manager Reports Error Message if it Can't Get Collection State
One of the following error messages may appear in the Remote Manager log:
msg_proc: e : Failure getting current collection states. Ignoring process ACMS01ACC001000 |
This message can be safely ignored. At the time this message is generated, the Remote Manager is attempting to synchronize collection states with the ACC. Because the ACC loads the Collection table, synchronization is implicit.
procmon: e : Failure obtaining current collection states. Bypassing <entity> |
In this error message, <entity> is an ACMS entity type.
You can disregard this message. At the time this message is generated,
the Remote Manager is attempting to synchronize collection states with
the process, but the process has either not fully initialized or is not
started. An example of this condition is when the ACMS run-time system
is started without the Queued Task Initiator (QTI). If the process is
initializing, synchronization will occur when the process finishes
initializing. If the process is not started, synchronization is not
necessary.
5.1.5 Remote Manager May Report ACMSTRC$_MONPROCEXST Error When Trying to Start Trace Monitor
In some situations, when the Remote Manager is trying to start the Trace monitor, it reports the following in the Remote Manager log:
rpc: e : %ACMSTRC-W-MONPROCEXST, ACMS$TRACE_MON process seems to already exist since ACMS$TRACE_MBX is defined |
This error message indicates that the Remote Manager attempted to create a Trace monitor process and detected that the logical name for the Trace monitor mailbox ACMS$TRACE_MBX was defined in the system logical name table. The process is assumed to exist if this logical name is defined. If there is a working version of the Trace monitor process running (image name ACMS$TRACE_MONITOR, process name ACMS$TRACE_MON), and the logical name ACMS$TRACE_MBX is defined as a mailbox device, then no action is required.
If the monitor process is malfunctioning, then you can try to stop it with the DCL command STOP/ID. If this fails, you can take the following steps:
The Remote Manager can then create a new process. If these attempts to
disable the existing Trace monitor process fail, then you should report
the problem to your Compaq support representative.
5.1.6 Remote Manager Takes a Few Seconds to Start
It takes a few seconds for the Remote Manager to become fully initialized during startup. This usually occurs on machines that are heavily loaded, on networks that are busy, or in command procedures that execute ACMSMGR commands immediately after starting the Remote Manager. During the Remote Manager initialization period, ACMSMGR commands are likely to fail with the following error messages:
%ACMSMGMT-W-NOCLNT_ATTACH, Cannot create client for node SPARKS %ACMSMGMT-E-NOCLIENTS, No clients created, cannot continue %ACMSMGMT-E-FAIL, Operation failed |
If this occurs, resubmit the command after pausing slightly. You can use the following series of commands in a command procedure to avoid this problem:
$ @SYS$STARTUP:ACMS$MGMT_STARTUP $ WAIT 00:00:05 $ ACMSMGR SHO ACC/ID |
The /SINCE qualifier of the ACMSMGR SHOW LOG command defaults to the current date only if the qualifier is specified as /SINCE= (that is, if there is an equal sign but no value string).
In addition, this command accepts only full and partial times and
dates. It does not accept the traditional OpenVMS generic qualifiers of
/TODAY or /YESTERDAY.
5.1.8 Remote Manager EVENT_SEVERITY Traps Do Not Include Text Messages
All Remote Manager traps report the date and time, severity, entity type, process name, trap parameter and current value. Remote Manager traps do not report text messages.
For event_severity traps, the traps include the severity of the event but do not indicate the actual text message. The traps receiver will know that a warning, an error, or a fatal event occurred, but it will not know what actually happened.
To determine which event caused the traps, enter the ACMSMGR SHOW LOG command and specify the traps time in the /SINCE qualifier.
For example, assume the following trap was received by a trap listener:
Node SPARKS Date/Time 5-OCT-1999 14:50:05.13 Severity Error Entity Type MGR Process Name ACMS$MGMT_SVR Trap Parameter event_s Current Value 2 |
This trap indicates that the Remote Manager on node SPARKS detected a warning level condition (value 2) at time 14:50:05.13. To determine the event that raised this condition, use the ACMSMGR SHOW LOG command as follows:
$ ACMSMGR SHO LOG/NODE=SPARKS/SINCE="5-OCT-1999 14:50"/SEV=WARN |
Note that the time of the event may slightly precede the trap time,
since the trap time is the time the trap was generated, not the time of
the event.
5.2 Restrictions Continued from ACMS Version 4.1
The following restrictions have been in effect since ACMS Version 4.1.
5.2.1 Form Name Cannot be a Logical Name When Using Multiple Submitter Platforms
When the ACMS$MULTIPLE_SUBMITTER_PLATFORMS is defined as "T", "Y" or 1, the ACMS EXC process attempts to open the form files specified in the task group definition by adding _VAX or _AXP to the file name. If the task-group definition uses a logical name for the form name, the EXC will not be able to locate the forms because no logical name translation is performed.
Form names when used in a multiplatform environment cannot be logical
names. The ACMS Version 4.3 user documentation reflects information
provided in this release note.
5.2.2 CDD Restriction Initializing Fields with Strings Shorter Than Field Size
Do not define a CDD field initialization string that is shorter than the field size. This may cause a memory access violation when using the Application Definition Utility (ADU) to build a group.
For example, the following CDO file sometimes causes ADU to generate a memory access violation when the ADU REPLACE GROUP statement is issued:
define field wksp_rep:r_1_field_1
datatype text size 30000
initial_value is "".
define record wksp_rep:r_1.
wksp_rep:r_1_field_1.
end record.
define field wksp_rep:r_2_field_1
datatype text size 30000
initial_value is "".
define record wksp_rep:r_2.
wksp_rep:r_2_field_1.
end record.
|
The workaround for this example is to remove the initial_value clauses.
5.2.3 Lowercase Characters Corrupted
If the terminal or terminal emulator is set to VT100 mode and both TDMS
and DECforms are present on the same system, lowercase characters in
DECforms menus are corrupted when an application is entered by
bypassing the ACMS menus. As a workaround, issue a SET TERMINAL
/DEC_CRT=2. This restriction occurs in ACMS Version 3.3 and higher.
5.2.4 On OpenVMS Alpha, Agent Hangs on Second Task Startup in Task Debugger
When an agent attempts to start a task in the ACMS Task Debugger while the Task Debugger is already executing a task, the agent should abort with the following error message:
%ACMS-E-EXCMAXDBGTSK, Only a single task may be debugged at one time |
As of ACMS Version 4.0 on OpenVMS Alpha, the agent hangs and does not
return the error message. This is due to a problem in OpenVMS Alpha
Version 6.x with certain $GETJPIW calls.
5.2.5 QTI's PIOPAGE Requirements for RMS Journaling on OpenVMS Alpha
On OpenVMS Alpha, most recovery unit-journaled $PUT, $UPDATE, or $DELETE operations require at least 8.2 pagelets (512-byte pages) of PIOPAGES for the duration of these operations. In addition, a typical call to SYS$END_TRANS or SYS$ABORT_TRANS to end a recovery unit may require approximately 16.5 pagelets of PIOPAGES for the duration of these operations. These figures are approximate and may vary depending on the characteristics of your application.
While the need for additional PIOPAGES is temporary over the duration
of the operation, RU applications that make extensive use of
asynchronous RMS operations may have several operations launched at the
same time, thus making heavier demands on their PIOPAGES. This is true
in the multithreaded QTI process. The calculation of PIOPAGES
requirements should be made by estimating the maximum number of threads
that will be used by the QTI and multiplying that number by 16.5.
5.2.6 Using the Terminal User Command SELECT
If a user is at the ACMS command menu and issues a SELECT command, the ACMS Command Process (CP) saves the selection and does not submit the task until the user issues a CONTINUE command. To use the SELECT command, you can do either of the following:
You cannot use the OpenVMS Phone utility (PHONE) from a DCL server when using ACMS because a user who enters ACMS using the ACMS/ENTER/NORETURN command or through a controlled terminal does not have an OpenVMS process associated with the user.
The Phone utility was designed for communication between interactive OpenVMS processes. DCL server processes are not true interactive OpenVMS processes. When the Phone utility tries to dial a user or answer a call, it checks the system for both an interactive process that has the correct user name and a login device that has the correct attributes.
When a user enters ACMS using the ACMS/ENTER command, the user's
original interactive process remains. PHONE sees the process and works
from the DCL server to which the user is attached. If the user enters
ACMS using the ACMS/ENTER/NORETURN command or through an ACMS
controlled terminal, there is no interactive process to meet the Phone
utility's requirements. Thus, the Phone utility does not work from the
DCL server to which the user is attached.
5.2.8 Submitters on VAXstation Must Have SHARE Privilege
A submitter logged in to a VAXstation device (WT or TK) cannot select a task that passes the terminal to a server process unless the server process has OpenVMS SHARE privilege. If the SHARE privilege is not defined, the submitter receives the %ACMSTRM-F-NOSHRPRV error and an entry is logged to the Software Event Logger (SWL).
SHARE privilege allows the server process to assign channels to devices
allocated to other users, such as a workstation device.
5.2.9 Task Debugger and Online Server Debugging
The following restrictions apply to the Task Debugger and online server
debugging.
5.2.9.1 Examining Binary Date Data Types
The ACMS Task Debugger EXAMINE command misrepresents the value of an OpenVMS ADT (quadword binary date) data type under some conditions, and accurately displays its contents under other conditions. For example, the following command format displays incorrect results:
ACMSDBG> EXAMINE <workspace-name> |
The workaround is to examine any binary date fields as individual, fully qualified single field names, rather than as part of the entire workspace. The following command displays the correct value for the date field:
ACMSDBG> EXAMINE/DATE <the-adt-date-field> OF <workspace-name> |
Exiting the ACMS Task Debugger with a Ctrl/C or Ctrl/Y may not delete
the subprocess created for a server. This can happen if Ctrl/C or
Ctrl/Y is used to exit the Task Debugger, and a server is attached to
an Rdb database and is waiting for access to the database.
5.2.10 Application Management Commands and Tools
The following restrictions apply to the application management tools
and commands.
5.2.10.1 ACMSPARAM May Generate an Excessive PGFLQUOTA Value for the CP Account
The current formula for calculating the PGFLQUOTA value of the CP account is liberal. The formula assumes that each exchange step involves the sending or receiving of the LARGEST_MESSAGE (see the Variables Required for ACMSPARAM.COM table in the Compaq ACMS for OpenVMS Managing Applications manual.
After executing ACMSPARAM, review the value of CP_PGFLQUOTA. If this value appears to be excessively large, then replace the LARGEST_MESSAGE variable with the average size of workspaces used in exchange steps, or with the size of the workspace most often used in exchange steps.
The formula for calculating the CP_PGFLQUOTA provides only a rough
approximation of this quota. To fine-tune your system, calculate the
quota for your particular system. The ACMS Version 4.2 user
documentation reflects information provided in this release note.
5.2.10.2 ACMSPARAM and ACMEXCPAR Do Not Support Search Lists
ACMSPARAM.COM and ACMEXCPAR.COM do not support logical name search
lists for the TDB file specifications.
5.2.10.3 MSS Process Pool Exhausted Errors are Reported Incorrectly
If a process runs out of ACMS message switch subsystem (MSS) process pool when allocating a buffer prior to reading a message from a network link, the status is truncated and the following error is written to the software error log:
%ACMSMSS-I-ERRLNKREAD, Error reading network link Unknown message - error code 0000A342 |
The correct reason is:
ACMSMSS-E-PROPOOEXH, Process local pool is exhausted |
To correct this problem, increase the MSS_PROCESS_POOL parameter in
ACMSGEN, and restart the ACMS system.
5.2.11 Task Queuing Restrictions
The following restrictions apply to task queuing.
5.2.11.1 Processing of Failed Queued Task Elements
If you are using a task queue file marked for RMS recovery-unit journaling, and a large number of queued task instances fail at the same time, in rare circumstances the QTI is unable to correctly process a failed queued task element. When this happens, the QTI cannot set the queued task element into a retry state, move it on to an error queue, or delete the queued task element. Instead, the failed queued task element is processed again immediately.
The QTI logs the following error entry in the ACMS audit log when this error occurs:
************************************************************ Type : ERROR Time : 14-OCT-1996 10:35:20.22 Queue : QLT$QUEUE1 ErrQue : QLT$ERRQUEUE1 Appl : QLT_APPL Task : QLT_TASK_X User : QLT_USER_42 Elem Id: 00000309-0000007E-D032A4E0-00948E68 Text : Error processing queued task -ACMSQUE-E-ERRPRCREC, Error trying to read the failed queued task record during error processing -RMS-S-OK_RRL, record locked against read but read anyway ************************************************************ |
| Previous | Next | Contents |