Skip to content

G
Gpdb

Greenplum / Gpdb

通知 7
Star 1
Fork 0
G
Gpdb

体验新版 GitCode，发现更多精彩内容 >>

提交 a1110fc7 编写于 作者: C Chuck Litzell 提交者： GitHub
浏览文件
操作

gpperfmon overview improvements (#2785) 

* gpperfmon overview improvements

* Add a link to the log rotation section.

* Edits from review
上级 23e5a5ee
隐藏空白更改
内联 并排
Showing with 202 addition and 99 deletion
gpdb-doc/dita/ref_guide/gpperfmon/db-database.xml
浏览文件 @ a1110fc7
......@@ -22,8 +22,7 @@
<li>
<codeph>database_history</codeph> is a regular table that stores historical
database-wide query workload data. It is pre-partitioned into monthly partitions.
Partitions are automatically added in two month increments as needed. Administrators
must drop old partitions for the months that are no longer needed.</li>
Partitions are automatically added in two month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/db-diskspace.xml
浏览文件 @ a1110fc7
......@@ -10,7 +10,7 @@
<codeph>diskspace_now</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. Current diskspace metrics are
stored in <codeph>database_now</codeph> during the period between data collection from
the Command Center agents and automatic commitment to the
the <codeph>gpperfmon</codeph> agents and automatic commitment to the
<codeph>diskspace_history</codeph> table.</li>
<li>
<codeph>diskspace_tail</codeph> is an external table whose data files are stored in
......@@ -21,8 +21,7 @@
<li>
<codeph>diskspace_history</codeph> is a regular table that stores historical diskspace
metrics. It is pre-partitioned into monthly partitions. Partitions are automatically
added in two month increments as needed. Administrators must drop old partitions for the
months that are no longer needed.</li>
added in two month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/db-filerep.xml
浏览文件 @ a1110fc7
......@@ -23,8 +23,7 @@
<li>
<codeph>filerep_history</codeph> is a regular table that stores historical database-wide
file replication data. It is pre-partitioned into monthly partitions. Partitions are
automatically added in two month increments as needed. Administrators must drop old
partitions for the months that are no longer needed.</li>
automatically added in two month increments as needed.</li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/db-interface-stats.xml
浏览文件 @ a1110fc7
......@@ -22,8 +22,7 @@
<li>
<codeph>interface_stats_history</codeph> is a regular table that stores statistical
interface metrics. It is pre-partitioned into monthly partitions. Partitions are
automatically added in one month increments as needed. Administrators must drop old
partitions for the months that are no longer needed.</li>
automatically added in one month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/db-log-alert.xml
浏览文件 @ a1110fc7
......@@ -5,23 +5,26 @@
<title> log_alert_* </title>
<body>
<p>The <codeph>log_alert_*</codeph> tables store <codeph>pg_log</codeph> errors and warnings. </p>
<p>See <xref href="dbref.xml#overview/section_ok2_wd1_41b"/> for information about configuring
the system logger for <codeph>gpperfmon</codeph>.</p>
<p>There are three <codeph>log_alert</codeph> tables, all having the same columns:</p>
<ul>
<li><codeph>log_alert_now</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. Current
<codeph>pg_log</codeph> errors and warnings data is stored in
<li><codeph>log_alert_now</codeph> is an external table whose data is stored in
<codeph>.csv</codeph> files in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/logs</codeph> directory. Current
<codeph>pg_log</codeph> errors and warnings data are available in
<codeph>log_alert_now</codeph> during the period between data collection from the
Command Center agents and automatic commitment to the <codeph>log_alert_history</codeph>
table.</li>
<li><codeph>log_alert_tail</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. This is a transitional table
for query workload data that has been cleared from <codeph>log_alert_now</codeph> but
has not yet been committed to <codeph>log_alert_history</codeph>. It typically only
contains a few minutes worth of data.</li>
<codeph>gpperfmon</codeph> agents and automatic commitment to the
<codeph>log_alert_history</codeph> table.</li>
<li><codeph>log_alert_tail</codeph> is an external table with data stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/logs/alert_log_stage</codeph>. This is a
transitional table for data that has been cleared from <codeph>log_alert_now</codeph>
but has not yet been committed to <codeph>log_alert_history</codeph>. The table includes
records from all alert logs except the most recent. It typically contains only a few
minutes' worth of data.</li>
<li><codeph>log_alert_history</codeph> is a regular table that stores historical
database-wide errors and warnings data. It is pre-partitioned into monthly partitions.
Partitions are automatically added in two month increments as needed. Administrators
must drop old partitions for the months that are no longer needed.</li>
Partitions are automatically added in two month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......@@ -246,29 +249,5 @@
</tbody>
</tgroup>
</table>
<section id="rotation">
<title>Log Processing and Rotation</title>
<p>The Greenplum Database system logger writes alert logs in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/logs</codeph> directory.</p>
<p>The agent process (<codeph>gpmmon</codeph>) performs the following steps to consolidate
log files and load them into the <codeph>gpperfmon</codeph> database:</p>
<ol>
<li>Gathers all of the <codeph>gpdb-alert-*</codeph> files in the logs directory (except
the latest, which the syslogger has open and is writing to) into a single file,
<codeph>alert_log_stage</codeph>.</li>
<li>Loads the <codeph>alert_log_stage</codeph> file into the
<codeph>log_alert_history</codeph> table in the <codeph>gpperfmon</codeph>
database.</li>
<li>Truncates the <codeph>alert_log_stage</codeph> file.</li>
<li>Removes all of the <codeph>gp-alert-*</codeph> files, except the latest.</li>
</ol>
<p>The syslogger rotates the alert log every 24 hours or when the current log file reaches
or exceeds 1MB. A rotated log file can exceed 1MB if a single error message contains a
large SQL statement or a large stack trace. Also, the syslogger processes error messages
in chunks, with a separate chunk for each logging process. The size of a chunk is
OS-dependent; on Red Hat Enterprise Linux, for example, it is 4096 bytes. If many
Greenplum Database sessions generate error messages at the same time, the log file can
grow significantly before its size is checked and log rotation is triggered.</p>
</section>
</body>
</topic>
gpdb-doc/dita/ref_guide/gpperfmon/db-queries.xml
浏览文件 @ a1110fc7
......@@ -13,7 +13,7 @@
<codeph>queries_now</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. Current query status is
stored in <codeph>queries_now</codeph> during the period between data collection from
the Command Center agents and automatic commitment to the
the <codeph>gpperfmon</codeph> agents and automatic commitment to the
<codeph>queries_history</codeph> table.</li>
<li>
<codeph>queries_tail</codeph> is an external table whose data files are stored in
......@@ -24,8 +24,7 @@
<li>
<codeph>queries_history</codeph> is a regular table that stores historical query status
data. It is pre-partitioned into monthly partitions. Partitions are automatically added
in two month increments as needed. Administrators must drop old partitions for the
months that are no longer needed.</li>
in two month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......@@ -137,10 +136,10 @@
<p>CPU usage by all processes across all segments executing this query (in
seconds). It is the sum of the CPU usage values taken from all active
primary segments in the database system.</p>
<p>Note that the value is logged as 0 if the query runtime
is shorter than the value for the quantum. This occurs even if the query
runtime is greater than the value for <codeph>min_query_time</codeph>,
and this value is lower than the value for the quantum.</p>
<p>Note that the value is logged as 0 if the query runtime is shorter than the
value for the quantum. This occurs even if the query runtime is greater than
the value for <codeph>min_query_time</codeph>, and this value is lower than
the value for the quantum.</p>
</entry>
</row>
<row>
......@@ -164,8 +163,8 @@
<p>Displays the amount of processing skew in the system for this query.
Processing/CPU skew occurs when one segment performs a disproportionate
amount of processing for a query. This value is the coefficient of variation
in the CPU% metric across all segments for this query,
multiplied by 100. For example, a value of .95 is shown as 95.</p>
in the CPU% metric across all segments for this query, multiplied by 100.
For example, a value of .95 is shown as 95.</p>
</entry>
</row>
<row>
......@@ -175,9 +174,9 @@
<entry>float</entry>
<entry>Displays the amount of row skew in the system. Row skew occurs when one
segment produces a disproportionate number of rows for a query. This value is
the coefficient of variation for the <codeph>rows_in</codeph> metric
across all segments for this query, multiplied by 100. For example, a
value of .95 is shown as 95.</entry>
the coefficient of variation for the <codeph>rows_in</codeph> metric across all
segments for this query, multiplied by 100. For example, a value of .95 is
shown as 95.</entry>
</row>
<row>
<entry>
......
gpdb-doc/dita/ref_guide/gpperfmon/db-segment.xml
浏览文件 @ a1110fc7
......@@ -18,7 +18,7 @@
<codeph>segment_now</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. Current memory allocation
data is stored in <codeph>segment_now</codeph> during the period between data collection
from the Command Center agents and automatic commitment to the segment_history
from the <codeph>gpperfmon</codeph> agents and automatic commitment to the segment_history
table.</li>
<li>
<codeph>segment_tail</codeph> is an external table whose data files are stored in
......@@ -29,8 +29,7 @@
<li>
<codeph>segment_history</codeph> is a regular table that stores historical memory
allocation metrics. It is pre-partitioned into monthly partitions. Partitions are
automatically added in two month increments as needed. Administrators must drop old
partitions for the months that are no longer needed.</li>
automatically added in two month increments as needed. </li>
</ul>
<p>A particular segment instance is identified by its <codeph>hostname</codeph> and
<codeph>dbid</codeph> (the unique segment identifier as per the
......
gpdb-doc/dita/ref_guide/gpperfmon/db-socket-stats.xml
浏览文件 @ a1110fc7
......@@ -22,8 +22,7 @@
<li>
<codeph>socket_stats_history</codeph> is a regular table that stores historical socket
statistical metrics. It is pre-partitioned into monthly partitions. Partitions are
automatically added in two month increments as needed. Administrators must drop old
partitions for the months that are no longer needed.</li>
automatically added in two month increments as needed. </li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/db-system.xml
浏览文件 @ a1110fc7
......@@ -11,7 +11,7 @@
<codeph>system_now</codeph> is an external table whose data files are stored in
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph>. Current system utilization
data is stored in <codeph>system_now</codeph> during the period between data collection
from the Command Center agents and automatic commitment to the
from the <codeph>gpperfmon</codeph> agents and automatic commitment to the
<codeph>system_history</codeph> table.</li>
<li>
<codeph>system_tail</codeph> is an external table whose data files are stored in
......@@ -22,8 +22,7 @@
<li>
<codeph>system_history</codeph> is a regular table that stores historical system
utilization metrics. It is pre-partitioned into monthly partitions. Partitions are
automatically added in two month increments as needed. Administrators must drop old
partitions for the months that are no longer needed.</li>
automatically added in two month increments as needed.</li>
</ul>
<table>
<tgroup cols="2">
......
gpdb-doc/dita/ref_guide/gpperfmon/dbref.xml
浏览文件 @ a1110fc7
......@@ -4,26 +4,35 @@
<title>The gpperfmon Database</title>
<body>
<p>The <codeph>gpperfmon</codeph> database is a dedicated database where data collection agents
on Greenplum segment hosts save statistics. The optional Greenplum Command Center management
tool requires the database. The <codeph>gpperfmon</codeph> database is created using the
on Greenplum segment hosts save query and system statistics. <ph otherprops="pivotal">The
optional Greenplum Command Center management tool depends upon the
<codeph>gpperfmon</codeph> database.</ph></p>
<p>The <codeph>gpperfmon</codeph> database is created using the
<codeph>gpperfmon_install</codeph> command-line utility. The utility creates the database
and the <codeph>gpmon</codeph> database role and enables monitoring agents on the segment
hosts. See the <codeph>gpperfmon_install</codeph> reference in the <cite>Greenplum Database
Utility Guide</cite> for information about using the utility and configuring the data
collection agents.</p>
<p>The <codeph>gpperfmon</codeph> database consists of three sets of tables.</p>
and the <codeph>gpmon</codeph> database role and enables the data collection agents on the
master and segment hosts. See the <codeph>gpperfmon_install</codeph> reference in the
<cite>Greenplum Database Utility Guide</cite> for information about using the utility and
configuring the data collection agents.</p>
<p>The <codeph>gpperfmon</codeph> database consists of three sets of tables that capture query
and system status information at different stages.</p>
<ul id="ul_oqn_xsy_tz">
<li>
<codeph>now</codeph> tables store data on current system metrics such as active
queries.</li>
<li><codeph>history</codeph> tables store data on historical metrics.</li>
<li><codeph>tail</codeph> tables are for data in transition. <codeph>Tail</codeph> tables are
for internal use only and should not be queried by users. The <codeph>now</codeph> and
<codeph>tail</codeph> data are stored as text files on the master host file system, and
accessed by the <codeph>gpperfmon</codeph> database via external tables. The
<codeph>history</codeph> tables are regular database tables stored within the
<codeph>gpperfmon</codeph>database.</li>
<li><codeph>_now</codeph> tables store current system metrics such as active queries. </li>
<li><codeph>_tail</codeph> tables are used to stage data before it is saved to the
<codeph>_history</codeph> tables. The <codeph>_tail</codeph> tables are for internal use
only and not to be queried by users.</li>
<li><codeph>_history</codeph> tables store historical metrics. </li>
</ul>
<p>The data for <codeph>_now</codeph> and <codeph>_tail</codeph> tables are stored as text files
on the master host file system, and are accessed in the <codeph>gpperfmon</codeph> database
via external tables. The <codeph>history</codeph> tables are regular heap database tables in
the <codeph>gpperfmon</codeph> database. History is saved only for queries that run for a
minimum number of seconds, 20 by default. You can set this threshold to another value by
setting the <codeph>min_query_time</codeph> parameter in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/conf/gpperfmon.conf</codeph> configuration file.
Setting the value to 0 saves history for all queries. </p>
<p>The <codeph>history</codeph> tables are partitioned by month. See <xref
href="#overview/section_et2_wmt_n1b" format="dita"/> for information about removing old
partitions.</p>
<p>The database contains the following categories of tables:</p>
<ul>
<li>The <codeph><xref href="db-database.xml#db-database">database_*</xref></codeph> tables
......@@ -32,15 +41,10 @@
store diskspace metrics.</li>
<li>The <codeph><xref href="db-filerep.xml#db-filerep">filerep_*</xref></codeph> tables store
health and status metrics for the file replication process. This process is how
high-availability/mirroring is achieved in Greenplum Database instance. Statistics are
high-availability/mirroring is achieved in a Greenplum Database instance. Statistics are
maintained for each primary-mirror pair.</li>
<li>The <codeph><xref href="db-interface-stats.xml#db-interface_stats"
>interface_stats_*</xref></codeph> tables store statistical metrics for each active
interface of a Greenplum Database instance. Note: These tables are in place for future use
and are not currently populated.</li>
<li>The <codeph><xref href="db-log-alert.xml#CommandCenterDatabaseReference-log_alert"
>log_alert_*</xref></codeph> tables store information about pg_log errors and
warnings.</li>
>log_alert_*</xref></codeph> tables store error and warning messages from pg_log.</li>
<li>The <codeph><xref href="db-queries.xml#db-queries">queries_*</xref></codeph> tables store
high-level query status information.</li>
<li>The <codeph><xref href="db-segment.xml#db-segment">segment_*</xref></codeph> tables store
......@@ -62,5 +66,128 @@
>memory_info</xref></codeph> view shows per-host memory information from the
<codeph>system_history</codeph> and <codeph>segment_history</codeph> tables.</li>
</ul>
<section id="section_et2_wmt_n1b">
<title>History Table Partition Retention</title>
<p>The <codeph>history</codeph> tables in the <codeph>gpperfmon</codeph> database are
partitioned by month. Partitions are automatically added in two month increments as needed. </p>
<p>The <codeph>partition_age</codeph> parameter in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/conf/gpperfmon.conf</codeph> file can be set to
the maximum number of monthly partitions to keep. Partitions older than the specified value
are removed automatically when new partitions are added. </p>
<p>The default value for <codeph>partition_age</codeph> is <codeph>0</codeph>, which means
that administrators must manually remove unneeded partitions.</p>
</section>
<section id="section_ok2_wd1_41b">
<title>Alert Log Processing and Log Rotation</title>
<p>When the <codeph>gp_gperfmon_enable</codeph> server configuration parameter is set to true,
the Greenplum Database syslogger writes alert messages to a <codeph>.csv</codeph> file in
the <codeph>$MASTER_DATA_DIRECTORY/gpperfmon/logs</codeph> directory. </p>
<p>The level of messages written to the log can be set to <codeph>none</codeph>,
<codeph>warning</codeph>, <codeph>error</codeph>, <codeph>fatal</codeph>, or
<codeph>panic</codeph> by setting the <codeph>gpperfmon_log_alert_level</codeph> server
configuration parameter in <codeph>postgresql.conf</codeph>. The default message level is
<codeph>warning</codeph>.</p>
<p>The directory where the log is written can be changed by setting the
<codeph>log_location</codeph> configuration variable in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/conf/gpperfmon.conf</codeph> configuration file. </p>
<p>The syslogger rotates the alert log every 24 hours or when the current log file reaches or
exceeds 1MB. </p>
<p>A rotated log file can exceed 1MB if a single error message contains a large SQL statement
or a large stack trace. Also, the syslogger processes error messages in chunks, with a
separate chunk for each logging process. The size of a chunk is OS-dependent; on Red Hat
Enterprise Linux, for example, it is 4096 bytes. If many Greenplum Database sessions
generate error messages at the same time, the log file can grow significantly before its
size is checked and log rotation is triggered.</p>
</section>
<section id="rotation">
<title>gpperfmon Data Collection Process</title>
<p>When Greenplum Database starts up with gpperfmon support enabled, it forks a
<codeph>gpmmon</codeph> agent process. <codeph>gpmmon</codeph> then starts a
<codeph>gpsmon</codeph> agent process on the master host and every segment host in the
Greenplum Database cluster. The Greenplum Database postmaster process monitors the
<codeph>gpmmon</codeph> process and restarts it if needed, and the <codeph>gpmmon</codeph>
process monitors and restarts <codeph>gpsmon</codeph> processes as needed.</p>
<p>The <codeph>gpmmon</codeph> process runs in a loop and at configurable intervals retrieves
data accumulated by the <codeph>gpsmon</codeph> processes, adds it to the data files for the
<codeph>_now</codeph> and <codeph>_tail</codeph> external database tables, and then into
the <codeph>_history</codeph> regular heap database tables. </p>
<note>The <codeph>log_alert</codeph> tables in the <codeph>gpperfmon</codeph> database follow
a different process, since alert messages are delivered by the Greenplum Database system
logger instead of through <codeph>gpsmon</codeph>. See <xref
href="#overview/section_ok2_wd1_41b" format="dita"/> for more information.</note>
<p>Two configuration parameters in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/conf/gpperfmon.conf</codeph> configuration file
control how often <codeph>gpmmon</codeph> activities are triggered:</p>
<ul id="ul_r1h_14b_41b">
<li>The <codeph>quantum</codeph> parameter is how frequently, in seconds,
<codeph>gpmmon</codeph> requests data from the <codeph>gpsmon</codeph> agents on the
segment hosts and adds retrieved data to the <codeph>_now</codeph> and
<codeph>_tail</codeph> external table data files. Valid values for the
<codeph>quantum</codeph> parameter are 10, 15, 20, 30, and 60. The default is 15.</li>
<li>The <codeph>harvest_interval</codeph> parameter is how frequently, in seconds, data in
the <codeph>_tail</codeph> tables is moved to the <codeph>_history</codeph> tables. The
<codeph>harvest_interval</codeph> must be at least 30. The default is 120. </li>
</ul>
<p>See the <codeph>gpperfmon_install</codeph> management utility reference in the
<cite>Greenplum Database Utility Guide</cite> for the complete list of gpperfmon
configuration parameters.</p>
</section>
<section>
<p>The following steps describe the flow of data from Greenplum Database into the
<codeph>gpperfmon</codeph> database when gpperfmon support is enabled.</p>
<ol id="ol_xcd_rbv_n1b">
<li>While executing queries, the Greenplum Database query dispatcher and query executor
processes send out query status messages in UDP datagrams. The
<codeph>gp_gpperfmon_send_interval</codeph> server configuration variable determines how
frequently the database sends these messages. The default is every second. </li>
<li>The <codeph>gpsmon</codeph> process on each host receives the UDP packets, consolidates
and summarizes the data they contain, and adds additional host metrics, such as CPU and
memory usage.</li>
<li>The <codeph>gpsmon</codeph> processes continue to accumulate data until they receive a
dump command from <codeph>gpmmon</codeph>.</li>
<li>The <codeph>gpsmon</codeph> processes respond to a dump command by sending their
accumulated status data and log alerts to a listening <codeph>gpmmon</codeph> event
handler thread.</li>
<li>The <codeph>gpmmon</codeph> event handler saves the metrics to <codeph>.txt</codeph>
files in the <codeph>$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph> directory on the
master host. </li>
</ol>
<p>At each <codeph>quantum</codeph> interval (15 seconds by default), <codeph>gpmmon</codeph>
performs the following steps:</p>
<ol id="ol_jvt_rzb_41b">
<li>Sends a dump command to the <codeph>gpsmon</codeph> processes.</li>
<li>Gathers and converts the <codeph>.txt</codeph> files saved in <codeph>the
$MASTER_DATA_DIRECTORY/gpperfmon/data</codeph> directory into <codeph>.dat</codeph>
external data files for the <codeph>_now</codeph> and <codeph>_tail</codeph> external
tables in the <codeph>gpperfmon</codeph> database. <p>For example, disk space metrics are
added to the <codeph>diskspace_now.dat</codeph> and <codeph>_diskspace_tail.dat</codeph>
delimited text files. These text files are accessed via the
<codeph>diskspace_now</codeph> and <codeph>_diskspace_tail</codeph> tables in the
<codeph>gpperfmon</codeph> database.</p></li>
</ol>
<p>At each <codeph>harvest_interval</codeph> (120 seconds by default), <codeph>gpmmon</codeph>
performs the following steps for each <codeph>_tail</codeph> file:<ol id="ol_elf_11c_41b">
<li>Renames the <codeph>_tail</codeph> file to a <codeph>_stage</codeph> file.</li>
<li>Creates a new <codeph>_tail</codeph> file.</li>
<li>Appends data from the <codeph>_stage</codeph> file into the <codeph>_tail</codeph>
file.</li>
<li>Runs a SQL command to insert the data from the <codeph>_tail</codeph> external table
into the corresponding <codeph>_history</codeph> table.<p>For example, the contents of
the <codeph>_database_tail</codeph> external table is inserted into the
<codeph>database_history</codeph> regular (heap) table.</p></li>
<li>Deletes the <codeph>_tail</codeph> file after its contents have been loaded into the
database table.</li>
<li>Gathers all of the <codeph>gpdb-alert-*.csv</codeph> files in the
<codeph>$MASTER_DATA_DIRECTORY/gpperfmon/logs</codeph> directory (except the most
recent, which the syslogger has open and is writing to) into a single file,
<codeph>alert_log_stage</codeph>.</li>
<li>Loads the <codeph>alert_log_stage</codeph> file into the
<codeph>log_alert_history</codeph> table in the <codeph>gpperfmon</codeph>
database.</li>
<li>Truncates the <codeph>alert_log_stage</codeph> file.</li>
</ol></p>
</section>
<p>The following topics describe the contents of the tables in the <codeph>gpperfmon</codeph>
database.</p>
</body>
</topic>
gpdb-doc/dita/utility_guide/admin_utilities/gpperfmon_install.xml
浏览文件 @ a1110fc7
......@@ -20,15 +20,14 @@
data collection agents. You must be the Greenplum Database system user
(<codeph>gpadmin</codeph>) to run this utility. The <codeph>--port</codeph> option is
required. When using the <codeph>--enable</codeph> option, the <codeph>--password</codeph>
option is also required. Use the <codeph>--port</codeph> option to
supply the port of the Greenplum Database master instance. If using the
<codeph>--enable</codeph> option, Greenplum Database
must be restarted after the utility completes.</p>
option is also required. Use the <codeph>--port</codeph> option to supply the port of the
Greenplum Database master instance. If using the <codeph>--enable</codeph> option, Greenplum
Database must be restarted after the utility completes.</p>
<p>When run without the <codeph>--enable</codeph> option, the utility just creates the
<codeph>gpperfmon</codeph> database (the database used to store system metrics collected
by the data collection agents). When run with the <codeph>--enable</codeph>
option, the utility also runs the following additional tasks
necessary to enable the performance monitor data collection agents:</p>
by the data collection agents). When run with the <codeph>--enable</codeph> option, the
utility also runs the following additional tasks necessary to enable the performance monitor
data collection agents:</p>
<ol>
<li id="ou143278">Creates the <codeph>gpmon</codeph> superuser role in Greenplum Database.
The data collection agents require this role to connect to the database and write their
......@@ -115,7 +114,7 @@ host all gpmon ::1/128 <b>password</b></codeblock></p>
<pt>--password <varname>gpmon_password</varname></pt>
<pd>Required if <codeph>--enable</codeph> is specified. Sets the password of the
<codeph>gpmon</codeph> superuser. Disallowed if <codeph>--enable</codeph> is not
specified.</pd>
specified.</pd>
</plentry>
<plentry>
<pt>--port <varname>gpdb_port</varname></pt>
......@@ -183,8 +182,8 @@ host all gpmon ::1/128 <b>password</b></codeblock></p>
</strow>
<strow>
<stentry>partition_age</stentry>
<stentry>The number of months that gperfmon statistics data will be retained. The default
it is 0, which means we won’t drop any data.</stentry>
<stentry>The number of months that gpperfmon statistics data will be retained. The default
value is 0, which means that partitions are never dropped automatically.</stentry>
</strow>
<strow>
<stentry>quantum</stentry>
......@@ -194,6 +193,12 @@ host all gpmon ::1/128 <b>password</b></codeblock></p>
amounts of data for system metrics, choose a higher quantum. To collect data more
frequently, choose a lower value.</p></stentry>
</strow>
<strow>
<stentry>harvest_interval</stentry>
<stentry>The time, in seconds, between data harvests. A data harvest moves recent data
from the <codeph>gpperfmon</codeph> external (<codeph>_tail</codeph>) tables to their
corresponding history files. The default is 120. The minimum value is 30. </stentry>
</strow>
<strow>
<stentry>ignore_qexec_packet</stentry>
<stentry>(Deprecated) When set to true, data collection agents do not collect performance
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或