products:promonitor:latest:monitorsguide:netweaver:icmmpiblocksandpipes

ICM MPI blocks and pipes monitoring

This monitor is dedicated to the memory pipes (MPI) of the Internet Communication Manager (ICM)
of an ABAP system. It reads, on every application server, how many MPI pipes and MPI blocks are
in use, compares them with their maximum and raises an alarm before the ICM runs out of them.
It also sends the counters as metrics, so the usage and its peak can be followed over time.

Note: MPI pipes carry the requests and responses exchanged between the ICM and the work processes, and MPI blocks are the shared memory buffers that hold their data. When either of them runs out, HTTP and other ICM requests are rejected or wait until memory is freed. Each application server has its own ICM and its own MPI memory, so the monitor reads them one by one: ICM_MPI_INFO for the counters, and report RSPFPAR for the profile parameter mpi/max_pipes. The counters are the values at the time of the run, except the peak, which is the highest block usage since the ICM started.

Prerequisites

  • The monitoring user needs the S_RFC authorization to call function module ICM_MPI_INFO.
  • The monitoring user needs to run report RSPFPAR through function module INST_EXECUTE_REPORT. The ABAP parameters monitor needs the same authorization.
  • The collector opens one RFC connection per application server, so each of them must be reachable from the collector, not only the one of the connector.
  • No transport is needed on the monitored system.

Configuration hints

  • Use the surveillance table to adjust the monitoring settings:
    • Thresholds
    • Severity
    • Filters
  • The application servers are read once per run and shared by all surveillance lines. Each line only filters them, so adding a line does not add any load on the system.
  • A new monitor has no surveillance line: no alarm is raised until a line is added. The metrics are sent anyway when Send metrics is enabled.
  • An application server that cannot be reached, or whose ICM does not answer, is skipped and the others are still reported. No alarm is raised for the skipped one: use the ICM status and usage monitor to be alerted when an ICM is not running.
  • The application servers are read one after the other within the timeout of the monitor. Raise the timeout on a system with many application servers.

Multi severity threshold

  • You can use the multi threshold syntax to define multiple threshold/severity settings per rule. The default thresholds use it: G2W:80 W2M:90 raises a WARNING alarm from 80% and a MAJOR alarm from 90%.

Global parameters

Parameter Default Description
Send metrics enabled If enabled, the metrics described below are sent to the configured plugins/integrations, for every application server, whether a surveillance line matches it or not.
Blocks pool All blocks Defines the pool of blocks on which Max blocks % is measured.
All blocks compares the blocks in use with the total number of blocks.
Non reserved blocks compares them with the general pool only, leaving the reserved blocks out.
Both percentages are always sent as metrics, so they can be compared before changing this setting. It only chooses the one the threshold reads.
When Non reserved blocks is selected and an application server does not report its non reserved block count, the threshold is measured on all blocks for that server, and the monitor log shows a warning.

Surveillance table

Parameter Description
Active Use this field to activate or deactivate a line of configuration.
Instance A filter to match only a given application server, such as appsrv_PRD_00.
Max pipes % The threshold for the MPI pipes in use, in percent of mpi/max_pipes. Either a multi severity threshold such as G2W:80 W2M:90, or a single value such as 80. An alarm is sent when the usage reaches this value. Leave at 0 to disable this check.
Max blocks % The threshold for the MPI blocks in use, in percent of the pool chosen in Blocks pool. Either a multi severity threshold such as G2W:80 W2M:90, or a single value such as 80. An alarm is sent when the usage reaches this value. Leave at 0 to disable this check.
Severity The severity of the alarm when the threshold is a single value. A multi severity threshold carries its own severities and ignores this field.
Auto clear If checked, the alarm is cleared automatically once its condition is no longer met.
Alarm tag A text used as a prefix of the alarm message of this line.
Alarm If checked, this line of surveillance will be used for alarm generation.

Note: every active line is evaluated on every application server it matches. An application server matched by two lines can raise the alarms of both lines.

Note: filters match any value containing the text, case insensitive. * matches any value and a leading ! excludes the values containing the text. To exclude several unrelated values, join them with +: !_00+!_01.

Note: a new line is created with * as instance filter and G2W:80 W2M:90 for both thresholds.

Note: when mpi/max_pipes is 0 on an application server, the pipes percentage cannot be computed. The pipes check is skipped for that server and the monitor log shows a warning.

Pipes maximum

  • The maximum number of pipes is not part of the counters. It is the profile parameter mpi/max_pipes, read with report RSPFPAR on each application server. The value set in the profile is used, or the system default when the profile does not set it.
  • The parameter only changes when the application server is restarted, and the report returns every profile parameter. The value is therefore kept by the monitor and read again once a day.
  • After a day, the value is refreshed on at most three application servers per run, and the previous value is used in the meantime. The first run reads every application server.
  • A changed mpi/max_pipes is therefore taken into account about a day after the restart, not at once.

Collection errors

  • The messages below are shown as the error of the monitor. The monitor log gives the details when the log level is set to debug: the answer of each function, the counters read and the percentages computed for each application server.
Message Description
Instance list is empty, failed to collect MPI usage The connector does not know the application servers of the system yet. They are read by the connector check.
No instance returned MPI usage : reason Every application server was skipped. The reason is the one of the last server: no connection, ICM not running, or an answer in which the counters were not found.
Report RSPFPAR did not return mpi/max_pipes on instance name : reason The pipes maximum could not be read: the report failed, did not list the parameter or returned a value that is not a number. The run fails, because the pipes usage cannot be computed without it.
The RFC authorization error returned by the system The monitoring user is not allowed to call ICM_MPI_INFO or INST_EXECUTE_REPORT. The monitor is flagged with an authorization error. See the prerequisites.

Note: a connection lost while the pipes maximum is being read does not fail the run. The previous value is used when there is one, otherwise that application server is skipped.

Alarm content

  • One alarm is raised per line, application server and check. The message contains the percentage, the number in use, the maximum and the application server:
    • ICM MPI : 85.0% pipes used (3400/4000) [appsrv_PRD_01]
    • ICM MPI : 81.35% blocks used (5200/6392) [appsrv_PRD_01]
  • With Blocks pool set to Non reserved blocks, the blocks alarm shows the non reserved pool:
    • ICM MPI : 90.39% non reserved blocks used (5200/5753) [appsrv_PRD_01]
  • In integrations that show the metric of an alarm, a pipes alarm is reported under ICM_MPI_PIPES_USED_PERCENT. A blocks alarm is reported under ICM_MPI_BLOCKS_USED_PERCENT, or under ICM_MPI_BLOCKS_USED_NOT_RESERVED_PERCENT when it is measured on the non reserved blocks.

Examples

Example 1

Active Instance Max pipes % Max blocks % Severity Auto clear Alarm tag Alarm
true * G2W:80 W2M:90 G2W:80 W2M:90 MAJOR true true

Effect: On each application server, sends a WARNING alarm when 80% of the pipes are in use and a MAJOR alarm from 90%, and the same for the blocks. The Severity field is not used, because both thresholds carry their own severities.

Example 2

Active Instance Max pipes % Max blocks % Severity Auto clear Alarm tag Alarm
true appsrv_PRD_00 70 0 CRITICAL true true
true !appsrv_PRD_00 G2W:80 W2M:90 G2W:85 W2M:95 MAJOR true true

Effect: The first line sends a CRITICAL alarm when application server appsrv_PRD_00 uses 70% or more of its pipes, and does not check its blocks. The second line watches the other application servers, with a WARNING alarm from 80% of the pipes or 85% of the blocks, and a MAJOR alarm from 90% of the pipes or 95% of the blocks.

Generated metrics

metricId metricUnit metricTarget metricDescription
ICM_MPI_PIPES_USED Pipes [INSTANCE] Sends the number of MPI pipes in use.
ICM_MPI_PIPES_LIMIT Pipes [INSTANCE] Sends the maximum number of MPI pipes, profile parameter mpi/max_pipes.
ICM_MPI_PIPES_USED_PERCENT Percent [INSTANCE] Sends the share of the pipes in use.
ICM_MPI_BLOCKS_USED Blocks [INSTANCE] Sends the number of MPI blocks in use.
ICM_MPI_BLOCKS_LIMIT Blocks [INSTANCE] Sends the total number of MPI blocks.
ICM_MPI_BLOCKS_RESERVED Blocks [INSTANCE] Sends the number of blocks of the reserved pool.
ICM_MPI_BLOCKS_NOT_RESERVED Blocks [INSTANCE] Sends the number of blocks of the general pool.
ICM_MPI_BLOCKS_PEAK Blocks [INSTANCE] Sends the highest number of blocks in use since the ICM started.
ICM_MPI_BLOCKS_USED_PERCENT Percent [INSTANCE] Sends the share of all blocks in use.
ICM_MPI_BLOCKS_USED_NOT_RESERVED_PERCENT Percent [INSTANCE] Sends the blocks in use as a share of the non reserved blocks.
ICM_MPI_BLOCKS_PEAK_PERCENT Percent [INSTANCE] Sends the peak as a share of all blocks.
ICM_MPI_TOTAL_SIZE_MB MB [INSTANCE] Sends the size of the MPI memory.

Note: the INSTANCE target holds the name of the application server.

Note: percentages and the size in MB are rounded to two decimals.

Note: the metrics are sent for every application server that answered, once per run, whatever the surveillance lines.

Note: a counter that the kernel does not report is not sent, rather than being sent as zero. This applies to the reserved and non reserved block counts, the peak and the total size. A percentage whose maximum is zero is not sent either.

products/promonitor/latest/monitorsguide/netweaver/icmmpiblocksandpipes.txt · Last modified: by rbariou