QRSUB(1) General Commands Manual QRSUB(1)
NAME
qrsub - Submit an Advance Reservation (AR) to Univa Grid Engine.
SYNTAX
qrsub [-a date_time] [-A account_string] [-ckpt ckpt_name] [-d time]
[-e date_time] [-fr yes|no] [-he yes|no] [-help] [-l resource_list]
[-now] [-M user[@host],...] [-N ar_name] [-P project_name] [-w e|v]
[-pe pe_name slot_range] [-q wc_queue_list] [-rao yes|no] [-si ses-
sion_id] [-u wc_user_list] [-cal_week weekly_calendar] [-cal_depth num-
ber] [-cal_jmp number]
DESCRIPTION
Qrsub provides a means for operators, managers or users referenced in
the ACL (see access_list(5)) "arusers" to create a Advance Reservation
(AR) or a Standing Reservation (SR) in the Univa Grid Engine queuing
system. ARs allow to reserve particular consumable resources for future
use. These reserved resources are only available for jobs requesting
the AR and the scheduler ensures the availability of the resources when
the start time is reached. Job requesting the AR can only use the
reserved consumable resources. SRs are recurring ARs which follow a
given calendar. All ARs within one SR has the same ID.
During AR submit time the Univa Grid Engine queuing system selects the
best suited queues for the AR request and then reserves the desired
amount of resources. For a reservation, all queues that are not in
orphaned state are considered as suited. Only if the AR request can be
fulfilled, the AR will be granted.
ARs will be deleted either automatically when the end time is reached,
or manually using qrdel . In both cases, first all jobs requesting the
AR will be removed and then AR itself. Already granted ARs can be shown
with qrstat(.)
Note: To make AR behavior predictable, it is necessary to have reserved
resources available at the time of AR start. This is done by keeping
jobs with a unlimited runtime limit separated from ARs, and not consid-
ering resources used by such jobs for reservation.
Note: Resource Quotas are not considered for AR queue selection and nor
for jobs requesting a AR.
When an AR was successfully added to the Univa Grid Engine queuing sys-
tem qrsub returns a unique integer ID referring the the newly created
AR. The highest AR ID is 9999999. If the highest ID is reached, a wrap-
around happens and the next unused ID, starting with 1, will be used.
For qrsub, the administrator and the user may define default request
files (analogous to Univa Grid Engine_request for qsub), which can con-
tain any of the possible command line options.
A cluster wide default request file is optional. If such a default
request file is used, it must be placed under
$SGE_ROOT/$SGE_CELL/common/sge_ar_request (global defaults file).
A user private default request file is optional. If it is used, it must
be placed under
$HOME/.sge_ar_request (user private defaults file).
OPTIONS
-a date_time
Defines the activation (start) date and time of an AR. The
option is not mandatory. If omitted, the current date_time is
assumed. Either a duration or end date_time must be specified
for ARs. For details about date_time please see sge_types(1) For
SRs the start time is optional and determines the earliest pos-
sible allocation time of an AR (but not the actual beginning of
the first AR).
-A account_string
Identifies the account to which the resource reservation of the
AR should be charged. For "account_string" value details please
see the "name" definition in sge_types(1). In the absence of
this parameter Univa Grid Engine will place the default account
string "sge" in the accounting record of the AR.
-cal_week weekly_calendar
If a calendar is defined it is a Standing Reservation request.
The weekly calendar has the same format as in the calendar
object (see calendar_conf(5)) with the exception that only "on"
as state is allowed. The weekly calendar defines the time ranges
for which AR instances should be created. The scheduler tries
to allocate a fixed amount of ARs. That amount can be influenced
with the depth ("-cal_depth") parameter. All ARs within a SR
have the same ID. Jobs running within one AR are deleted at AR
end (like with ARs) but queued jobs are staying queued waiting
for the next AR occurrence within the SR. Giving a start time,
end time, or duration is optional for SRs. If given the start
time determines the earliest allowed start time for the first
AR. The end time determines the last possible end time of the
last AR within the SR. The duration specifies the end time. If
one AR within the SR ends new ARs are scheduled automatically,
so that the total amount of allocated ARs are equal to the SR
depth.
-cal_depth number
Specifies the depth of the SR. The depth determines how many ARs
are allocated (scheduled) at one point in time for the future.
ARs which can not get enough resources are un-allocated and not
counted. If un-allocated ARs are allowed for the SR is deter-
mined by the "-cal_jmp" parameter. If one AR ends further ARs
are allocated so that the depth is restored again.
-cal_jmp number
The jmp parameter determines how many calendar entries can be
jumped over if they can not be allocated for ARs due to missing
resources. Per default it is not allowed to jump over non-allo-
cated ARs (0). The SR will end then.
-ckpt ckpt_name
Selects the checkpointing environment (see checkpoint(5)) the AR
jobs may request. Using this option guarantees queues that only
providing this checkpoint environment will be reserved.
-d time
Defines the duration of the AR. The use of "-d time" is optional
if "-e date_time" is requested. For details about "time" defi-
nition please see sge_types(1). For SRs the duration is
optional and determines implicitly the end time, i.e. the last
possible end time for the last AR within the SR.
-e date_time
Defines the end date and time of an AR. The use of "-e
date_time" is optional if "-d time" is requested. For details
about "date_time" definition please see sge_types(1). For SRs
the end time is optional and determines the last possible end
time for the last AR within the SR.
-fr y[es]|n[o]
Specifies the behaviour of the AR at AR start time. With ("-fr
yes") jobs still blocking resources being granted to the AR will
be killed. With default behaviour ("-fr no") jobs holding
resources being granted to the AR can continue running and may
prevent start of jobs in the AR. In the current implementation
only resources requested by the AR via exclusive complexes are
freed. The feature is suited for freeing whole queue instances
or hosts or even the whole cluster, e.g. for maintenance pur-
poses. Requesting "-fr yes" requires operator rights.
-he y[es]|n[o]
Specifies the behavior when the AR goes into an error state. The
AR goes into error state when a reserved host goes into unknown
state, a queue error happens, or when a queue is disabled or
suspended.
A hard error, "-he yes", means as long as the AR is in error
state no jobs using the AR will be scheduled. If soft error,
"-he no", is specified the AR stays usable with the remaining
resources.
By default soft error handling is used.
-help Prints a list of all options.
-l resource=value,...
Creates an AR in a Univa Grid Engine queue, providing the given
resource request list. complex(5) describes how a list of
available resources and their associated valid value specifiers
can be obtained.
There may be multiple -l switches in a single command.
-m b|e|a|n
Defines or redefines under which circumstances mail is to be
sent to the AR owner or to the users defined with the -M option
described below. The option arguments have the following mean-
ing:
`b' Mail is sent at the beginning of the AR
`e' Mail is sent at the end of the AR
`a' Mail is sent when the AR goes into error state
`n' No mail is sent, default for qrsub
-M user[@host],...
Defines or redefines the list of users to which the qmaster
sends mail.
-masterq wc_queue_list
Only meaningful for a parallel AR request together with the -pe
option.
This option is used to reserve the proper queues to match this
request if it would be requested by a qsub. A more detailed
description of wc_queue_list can be found in sge_types(1).
-now y[es]|n[o]
This options impacts the queues selection for reservation.
With the "-now y" option, only queues with the qtype "INTERAC-
TIVE" assigned will be considered for reservation. "-now n" is
the default for qrsub.
-N name
The name of the AR. The name, if requested, must conform to
"name" as defined in sge_types(1). Invalid names will be denied
at submit time.
-P project_name
This option binds the AR to a project. Only members of this
project are allowed to use this AR.
-w e|v Specifies the validation level applied to the AR request.
The specifiers e and v define the following validation modes:
`v' verify - does not submit the AR but prints
an extensive validation report
`e' error - rejects request if requirements cannot
be fulfilled, default for qrsub
-pe parallel_env n[-[m]]|[-]m,...
Parallel programming environment (PE) to select for the AR queue
reservation. Please see the details of a PE in sge_pe(5).
-q wc_queue_list
Defines or redefines a list of cluster queues, queue domains or
queue instances, that may be reserved by the AR. Please find a
description of wc_queue_list in sge_types(1). This parameter
has all the properties of a resource request and will be merged
with requirements derived from the -l option described above.
-rao y[es]|n[o]
Specifies which resources will be considered for the AR at sub-
mission / reservation time. With -rao yes only currently avail-
able resources will be reserved, this means only queues which
are not in disabled, suspended or error state. Default setting
for -rao is "no". A cluster wide setting for only reserving
available resources can be done via qmaster_param
AR_RESERVE_AVAILABLE_ONLY, see also sge_conf(5).
-si session_id
Requests sent by this client to the sge_qmaster(1) daemon will
be done as part of the specified session. If the switch is omit-
ted or if NONE is specified as session_id then such requests
will be executed outside the control of a session.
Find more information concerning sessions in session_conf(5).
-u [username|@access_list],...
Defines the users allowed to submit jobs requesting the AR. The
access is specified by a comma separated list containing UNIX
users or ACLs (see access_list(5)). prefixing the ACL name with
an '@' sign.
By default only the AR owner is allowed to submit jobs request-
ing the AR.
Note: Only queues, where all users specified in the list have
access, are considered for reservation (see queue_conf(5)).
ENVIRONMENTAL VARIABLES
SGE_ROOT Specifies the location of the Univa Grid Engine standard
configuration files.
SGE_CELL If set, specifies the default Univa Grid Engine cell. To
address a Univa Grid Engine cell, qsub, qsh, qlogin or
qalter use (in the order of precedence):
The name of the cell specified in the environment
variable SGE_CELL, if it is set.
The name of the default cell, i.e. default.
SGE_DEBUG_LEVEL
If set, specifies that debug information should be writ-
ten to stderr. In addition the level of detail in which
debug information is generated is defined.
SGE_QMASTER_PORT
If set, specifies the TCP port on which sge_qmaster(8)
is expected to listen for communication requests. Most
installations will use a services map entry for the ser-
vice "sge_qmaster" instead of defining the port.
FILES
$SGE_ROOT/$SGE_CELL/common/sge_ar_request
global defaults file
$HOME/.sge_ar_request
user private defaults file
SEE ALSO
qrdel(1), qrstat(1), qsub(1), sge_types(1), checkpoint(5), complex(5),
queue_conf(5), session_conf(5), sge_pe(5), sge_resource_quota(5). cal-
endar_conf(5).
COPYRIGHT
See sge_intro(1) for a full statement of rights and permissions.
Univa Grid Engine User Commands UGE 8.5.4 QRSUB(1)