p4 monitor
Display Perforce process information and control long-running tasks.
Syntax
p4 [g-opts] monitor show [-a -l -e -L -s R | T | P | B | F | I ]
p4 [g-opts] monitor terminate id
p4 [g-opts] monitor clear {id | all}
p4 [g-opts] monitor pause id
p4 [g-opts] monitor resume id p4 [g-opts] monitor realtime [-F -T]
Description
p4 monitor allows a system administrator to observe and control Helix Server-related processes running on a Helix Server machine. Processes are tracked using a dedicated table that is constantly updated. This has a minor impact on server performance.
To use p4 monitor
, you must enable monitoring on the
Perforce
service by setting the monitor
configurable with p4 configure
.
Valid values for the monitor configurable are:
- 0: Server process monitoring off. (Default)
- 1: monitor active commands
- 2: active commands and idle connections
- 3: same as
2
, but also includes connections that failed to initialize (stuck at the Init() phase) - 5: same as
2
, but also includes a list of the files locked by the command for more than one second - 10: same as
5
, but also includes lock wait times - 25: same as
10
, except that the list of files locked by the command includes files locked for any duration
Changes to the monitor
configurable affect all new
p4
processes that connect to the server. Restarting
the server is not required.
See Enabling process monitoring in Helix Core Server Administrator Guide.
Command syntax variants provide the following alternatives:
-
(
list
level access):To list current process information, usep4 monitor show
. By default, all processes are listed, but only the command (for example,sync
,edit
,submit
) is shown, without arguments. Use the-s status
option to restrict the display to processes in the specified state. - (
super
level access):To show the list of arguments associated with each command, use the-a
(arguments) option or-l
(long) option. For additional information from the user environment, use the-e
(environment) option. To show locked files, use the-L
option. -
(operator or
super
level access): To mark a process for termination, usep4 monitor terminate id
. This command requires that the user be an operator or havesuper
level access.The
p4 monitor terminate
command does not mark a process for termination unless the process has been running for at least ten seconds. Some commands, such asp4 obliterate
, cannot be terminated.To control how often the list of processes is refreshed, see the
db.monitor.interval
, which is also mentioned in the Support Knowledgebase article, "Fixing a hung Helix Server". -
(operator or
super
level access): If a command terminates prematurely on the server side, it might be erroneously listed as running. You can clear such processes withp4 monitor clear
.With
super
level access:- To remove an entry from the monitor table, use
p4 monitor clear id
- To clear the entire table, use
p4 monitor clear all
Processes marked as running continue to run to completion even if removed from the monitor table with
p4 monitor clear
. - To remove an entry from the monitor table, use
-
(operator or
super
level access): To control the following tasks if they are running too long, use thep4 monitor pause
andp4 monitor resume
:
Output format
Each line of p4 monitor
output consists of the
following fields:
pid
status
owner
hh:mm:ss
command
[args
]
pid |
The process ID under Unix (or thread ID under Windows) |
status |
R,
Note
Finish and Background occur only in replica servers. |
owner |
The Helix Server user name of the user who invoked the command. |
hh:mm:ss |
The time elapsed since the command was called. |
command [args] |
The command and arguments as received by the Perforce service. |
For example, consider the following output to the p4 monitor
show -L
command, which displays information about locked
files:
8764 R user 00:00:00 edit [server.locks/clients/88,d/ws4(W),db.locks(R),db.rev(R)] 8766 R user 00:00:00 edit [server.locks/clients/89,d/ws5(W),db.locks(R),db.rev(R)] 8768 R user 00:00:00 monitor
Following pid, status, owner, and time information, this shows two edit
commands that have various files locked, including the client workspace
lock in exclusive mode for the workspaces ws4
and
ws5
, and db.locks
and db.rev
tables in read-only mode.
If you have enabled idle process monitoring (by setting the
monitor
configurable to 2), idle processes appear with a
status
of R
, but with a
command
of IDLE
.
Some commands (for instance, p4
submit
) invoke multiple processes. For example,
dm_CommitSubmit
or dm_SubmitChange
might appear
in the output of p4 monitor
as two separate phases of
the p4 submit
command.
Getting pull thread information for replicas
If you are running a replica with monitoring enabled and you have not
configured the monitor table to be disk-resident, you can run the
following command to get more precise information about what pull threads
are doing. (Remember to set monitor.lsof
).
$ p4 monitor show -sB -la -L
Command output would look like this:
31701 B uservice-edge3 00:07:24 pull sleeping 1000 ms [server.locks/replica/49,d/pull(W)]
p4 monitor realtime
Outputs real-time performance counters when enabled by setting the rt.monitorfile configurable. One use case is to determine, in real time, whether a replica is making progress while replicating a large transaction from the master. (This use case requires that the rpl.track.behind configurable be set to 1
or 2
.)
p4 monitor realtime options
You must be an operator or have super
access to use the
following options:
-F
|
List only performance counters satisfying the filter expression. The filter evaluates the contents of the fields in the preceding list. Filtering is case-sensitive. Example: (The filter syntax is similar to the one used for Job Views.) |
|
Returns only the specified fields. The field names can be specified using a comma-delimited list or a space-delimited list. Example: |
The performance counters can be read at the command line or by external tools.
command line | external tools |
---|---|
Use any of the following:
See Helix Core Server (p4d) Reference and Monitoring the server in Helix Core Server Administrator Guide. |
Use the MonItems class in Helix Core C/C++ Developer Guide |
The real-time performance counters are for monitoring in real time:
Counter |
Description |
---|---|
rtv.db.ckp.active
|
Indicates if a checkpoint is in progress. |
rtv.db.ckp.records
|
The number of records that have been read into an ongoing checkpoint. |
rtv.db.io.records
|
Running count of database IO reads, writes, and deletes. This counter
might affect performance. To enable, set
the db.rt.io configurable to 1 . |
rtv.db.lockwait
|
Commands waiting for locks. Includes the high mark of most commands waiting for locks at any one time. |
rtv.rpl.behind.bytes
|
Replication lag represented as bytes of
journal that the upstream server has yet
to send. This counter is enabled when the
rpl.track.behind configurable is set to
1 or greater.
The rpl.track.behind.interval
configurable can be used to tune the
frequency of updates during p4 pull. |
rtv.rpl.behind.journals
|
Replication lag represented as rotated
journals that the upstream server has yet
to send.
See rtv.rpl.behind.bytes for related
configurables. |
rtv.server.connection.pause.pct.cpu
|
The percentage of commands to pause when CPU pressure is in a high state. The highest percentage of this or the |
rtv.server.connection.pause.pct.mem
|
The percentage of commands to
pause when memory pressure is in a medium
or high state. The highest percentage of
the |
rtv.svr.sessions.active
|
Current connections from clients. Includes a high mark of most concurrent connections at any one time. |
rtv.svr.sessions.paused
|
Number of server sessions that are paused. |
rtv.svr.sessions.total
|
Running count of client connections. |
rtv.sys.mem.os.available
|
An estimate of how much memory is available for starting new applications without swapping. |
rtv.sys.mem.os.free
|
Amount of memory that is not allocated to any application or the operating system. |
rtv.sys.mem.os.swap.free
|
Amount of swap memory that is not allocated. |
rtv.sys.mem.os.swap.total
|
Total amount of swap configured in the system. |
rtv.sys.mem.os.total
|
Total amount of operating system memory. |
rtv.sys.pressure.lastcollecttime
|
Timestamp of the most recent assessment of system pressure. |
rtv.sys.pressure.level.cpu |
The level of CPU pressure on a scale from 0 = low, 1 = medium, 2 = high |
rtv.sys.pressure.level.mem
|
The level of memory pressure on a scale from 0 = low, 1 = medium, 2 = high |
Options
You must be an operator or have super
access to use the
following options:
|
Show all arguments associated with the process (for example,
Helix Server user names are truncated to 10 characters, and each line of output is limited to 80 characters. |
|
Show environment information, including invoking Helix Server application (if known), host IP address, and workspace name. |
|
Show all arguments in long form, that is, without truncating user names or the list of command line arguments. |
|
Show information about locked files. The information is
collected only for the duration of the Pre-requisites for using this option vary with the platform on which the server is running.
You can use the |
|
Restrict the display to processes in the |
|
See Global options. |
Usage Notes
Can File Arguments Use Revision Specifier? | Can File Arguments Use Revision Range? | Minimal Access Level Required |
---|---|---|
No |
No |
available to an operator user
|
Examples
|
Show
Helix Server
process information (commands only). Requires |
|
Show arguments and commands, without limits on line length.
Requires |
|
Show arguments and commands, limited to 80 characters per line
of output. Requires |
|
Instruct the
Perforce
service to mark process 123 for termination. Requires
|
|
Clears the monitor table of all entries. Requires
|
Related Commands
To turn on monitoring |
|
To turn off monitoring |
|