vxtask - list and administer Veritas Volume Manager tasks
vxtask abort taskid
vxtask [-ahlpqr ] [-g dg_name [-g dg_name...]] [[-G dg_name] -v volume [-v volume...]] [-i taskid] list [taskid...]
vxtask pause taskid
vxtask resume taskid
vxtask [-i taskid] set name=value taskid
vxtask [-c count] [-t time] [-w wait_interval] [-ln ] monitor [taskid[taskid...]]
The vxtask utility performs basic administrative operations on Veritas Volume Manager (VxVM) tasks that are running on the system. Operations include listing tasks (subject to filtering), modifying the state of a task (aborting, pausing, resuming), and modifying the rate of progress of a task.
Veritas Volume Manager tasks represent long-term operations in progress on the system. Every task provides the time the operation started, the size and progress of the operation, and the operations state and rate of progress (throttle). You can change the state of a task, giving coarse-grained control over an operations progress. For those operations that support throttling, you can change the rate of progress of a task, giving more fine-grained control over the task.
The majority of tasks represent I/O being performed to objects. Operations such as read-writeback recovery for mirrored volumes, parity recalculation for RAID-5 volumes, volume relayouts, and so forth, all involve moderate to large amounts of I/O. Tasks that represent operations that perform I/O directly are referred to as basic tasks. The task progress information for basic tasks consists of the starting and ending block of the I/O to be performed and the offset to which the I/O has currently completed.
Some operations, such as vxrecover and starting RAID-5 volumes, may require multiple tasks to complete the operation. vxrecover may determine that many recoveries are required, some that must be performed serially. Starting a RAID-5 volume requires that its log (if any) be replayed and cleared, its parity be recalculated (if necessary), any stale subdisks be resynchronized (if needed), and that these tasks be performed serially. In these and similar cases, an overall task is created to keep track of the underlying basic tasks being performed, and these are known as parent tasks. The progress information for parent tasks are the total number of subtasks required to complete the overall operation; the number of subtasks completed; and the number of subtasks currently running.
Every task is given a unique task identifier. This is a numeric identifier for the task that can be specified to the vxtask utility to specifically identify a single task.
Tasks also contain the following information:
task tag A tasks tag is a string that you can specify to make administration easier. It is usually set by the command that initiates the task. For most utilities, the -t tasktag option allows a tag to be specified as a sequence of up to 16 alphanumeric characters. task type The tasks type describes the specific operation that is being performed. For example, attaching a plex to a volume results in an atomic copy loop, which is represented by an ATOMIC_COPY task. description Describes the utility operation on whose behalf the I/O is being performed. For example, both a volume start and a plex attach can result in an atomic copy loop and thus an ATOMIC_COPY task. The description of the operation attempts to disambiguate between tasks that are performing the same basic work. object Most tasks are related to a specific object, usually the volume within which the operation is taking place. parent Some utilities require multiple operations to fulfill a request. For example, the vxrecover utility may need to start many different volumes. In these cases, the subtasks list their parent. progress information Each task has an indication of its starting point and ending point, and an indication as to what point it has progressed so far. Combined with the work time, this allows the amount of time remaining until completion to be estimated for some tasks. state Each task has one of three states: RUNNING, PAUSED or ABORTING. time Each task notes its starting time and the total amount of time spent doing work. nodename For tasks that are being processed remotely by the use of the distribute keyword in /etc/default/vxrecover, the name of the cluster node on which the task is being processed is also stored.
abort|pause|resume These three operations request that the specified task change its state. pause puts a running task in the paused state, suspending its operation. resume continues the operation of a paused task. abort stops the operation of the specified. In most cases, the operations "back out" in a manner similar to when an I/O error occurs, reversing as much as possible whats been done so far. -H Provides full help for the vxtask utility; same as the help option. help Provides full help for the vxtask utility; same as the -H option. list List tasks running on the system in one-line summaries. The value displayed for percentage task progress takes into account all intermediate steps that are required to accomplish a specific task. By default, all tasks running on the system are printed. If a taskid argument is supplied, the output is limited to those tasks whose taskid or task tag match taskid. The remaining options filter and limit the listed tasks as follows:
-a Limits the output to tasks in the aborting state. -g dg_name Limits the listed tasks to those tasks running on objects in the diskgroup dg_name. -G Distinguishes between volumes with the same name in different diskgroups; see vxintro(1M). -h Prints tasks hierarchically with a tasks child tasks following the parent task. -l Prints tasks in long format. -p Limits the output to tasks in the paused state. -q Suppresses printing of column headings. -r Limits the output to tasks in the running state. -v volume Limits the output to tasks whose object is volume. Any task meeting any of the criteria specified are printed. That is, if -v foovol and -p are specified, then any task operating on foovol or that is paused are printed. If -h were also specified, then those tasks and all their children would be listed. monitor The monitor operation continuously prints information about a task or group of tasks as the task information changes. The -l option specifies printing a long listing. The default is to print one-line listings. The monitor command accepts the following options:
-c count Prints count sets of task information, then exits. -n Monitors not only the tasks specified on the command line, but also any tasks newly registered while the program is running. -t time Exits after time seconds. -w interval Prints the string "waiting..." when interval seconds pass with no output activity. In addition to printing task information when a task state changes, output is also generated when the task completes. When this occurs, the task state is printed as EXITED (see the OUTPUT section). set The set operation changes modifiable parameters of a task. There is only one modifiable parameter for tasks: the slow attribute, which represents a throttle on the task progress. The larger the slow value, the slower the progress of the task and the fewer system resources it consumes in a given time. (The slow attribute is the same attribute that many commands, such as vxplex, vxvol and vxrecover, accept.)
There are two output formats printed by vxtask: a short, one-line summary format per task, and a long task listing. The short listing attempts to provide the most used task information for quick perusal, displaying the following fields from left to right:
1. The task ID for the task. 2. The task ID of the tasks parent, if any. If the task has no parent, this field is left blank. 3. The task type and state, separated by a slash (/). The type field is a description of the work being performed, such as ATOMIC COPY for atomic copy I/O and RELAYOUT for an online relayout operation. The state is a single letter representing the task state: R for running, P for paused, and K for aborting. In the special case of the completion of a task that is being monitored, the state is printed as EXITING. 4. The percentage of the operation that has been completed to this point. 5. The tasks progress information. For basic tasks, this is the starting offset, ending offset and current offset for the operation, separated by slashes (/). For parent tasks, this is the starting number of child tasks, total number of child tasks required for the operation to complete, and the number of child tasks already completed to this point, separated by slashes, followed by the number of children currently running in parentheses. 6. A description of the command for which the task is performing work. This is normally a single word, such as START for a volume start operation or RELAYOUT for an online relayout, followed by object names, indicating what the arguments were for the command. See EXAMPLES below. 7. If the task is using the smartmove faeture to perform the admin operations (like sync, resync, refresh, reclaim etc), then the flag smartmove is displayed after the object information. 8. For tasks that are being processed remotely, the name of the cluster node on which the task is being processed is displayed at the end of the line within parentheses.
The long output format prints all available information for a task, spanning multiple lines. If more than one task is printed, the output for different tasks is separated by a single blank line. Each line in the long output format contains a title for the line, followed by a colon (:), followed by the information. The possible information printed for a task includes:
Operation: A description of the operation on whose behalf the work is being performed. This is essentially a description of the command executed along with the Veritas Volume Manager objects specified as arguments. Progress: A description of the progress of the task. The progress is stated as a percentage of completion, along with the actual progress values for the task. Started: The time that the task was started. Task: General information for the task. The tasks ID, followed by its parents ID (if any) in parentheses, followed by the tasks state, either RUNNING, PAUSED or ABORTING. Throttle The throttle value for the task, if applicable. Type: The task type, describing the work being performed. Work time: States how much time has been spent performing the work, and a rough estimate of the time remaining for the task to complete. Node: For remotely processed tasks, this field states the name of the cluster node on which the task is being processed. For locally processed tasks, this field is not displayed. As a special case, when a task being monitored completes only the Task: line is printed for the task, with the task state represented by EXITED.
To list all tasks currently running on the system:
To trace all tasks in the diskgroup foodg that are currently paused, as well as any tasks with the tag sysstart:
vxtask -g foodg -p -i sysstart list
To list all tasks on the system that are currently paused:
vxtask -p list
To monitor all tasks with the tag myoperation:
vxtask monitor myoperation
To cause all tasks tagged with recovall to exit:
vxtask abort recovall
/dev/vx/info Used to get task information from the kernel. /dev/vx/taskmon Task monitoring device.
vxintro(1M), vxplex(1M), vxrecover(1M), vxsd(1M), vxvol(1M)