GlideinWMS The Glidein-based Workflow Management System

WMS Factory

Custom HTCondor Variables

Description

This document describes what configuration variables are used by the glideins. Most administrators never need to touch most of them, but a sophisticated Glidein Factory administrator may need to tweak some of them to implement the desired policies (for example: require encryption over the wire) or to address the needs of a particular site (for example: max allowed wallclock time).

Configuration variable location

The GlideinWMS ships with a set of pre-defined configuration variables, that are stored in two files (known as condor vars files):

glideinWMS/creation/web_base/condor_vars.lst
glideinWMS/creation/web_base/condor_vars.lst.entry

The two files are equivalent, but were split for historical reasons, and the second one is meant to contain site specific configuration variables.
These files should never be modified, and represent just the default shipped by the software!

A GlideinWMS administrator can change the values of the predefined variables (with some exceptions, see below), and define new ones using the Glidein Factory configuration file.

HTCondor vars files

The condor vars files contain the GlideinWMS pre-defined configuration variables, and should never be modified.
However, a GlideinWMS administrator should nevertheless be able to read them.

Each of them is an ASCII file, with one entry per row.
Lines starting with # are comments and are ignored.

Each non comment line must have 7 columns. Each column has a specific meaning:

  1. Attribute name
  2. Attribute type
    • I (int) – integer
    • S (string) – quoted string
    • C (expr) – unquoted string (i.e. HTCondor keyword or expression)
  3. Default value, use – if no default
  4. HTCondor name, i.e. under which name should this attributed be known in the configuration used by HTCondor daemons
  5. Is a value required for this attribute?
    Must be Y or N. If Y and the attribute is not defined, the glidein will fail.
  6. Will condor_startd publish this attribute to the collector?
    Must be Y or N.
  7. Will the attribute be exported to the user job environment?
    • - - Do not export (for glidein/condor internal use)
    • + - Export to the user job environment using the original attribute name
    • @ - Export to the user job environment using the HTCondor name

Here below, you can see a short extract; the semantics of the variables is defined below.

# VarName               Type    Default         CondorName                      Req.    Export  UserJobEnvName
#                       S=Quote - = No Default  + = VarName                             HTCondor  - = Do not export
#                                                                                               + = Use VarName
#                                                                                               @ = Use CondorName
#################################################################################################################
X509_USER_PROXY         C       -               GSI_DAEMON_PROXY                Y       N       -
USE_MATCH_AUTH          C       -     SEC_ENABLE_MATCH_PASSWORD_AUTHENTICATION  N       N       -
GLIDEIN_Factory         S       -               +                               Y       Y       @
GLIDEIN_Name            S       -               +                               Y       Y       @
GLIDEIN_Collector       C       -               HEAD_NODE                       Y       N       -
GLIDEIN_Expose_Grid_Env C       False     JOB_INHERITS_STARTER_ENVIRONMENT      N       Y       +
TMP_DIR                 S       -               GLIDEIN_Tmp_Dir                 Y       Y       @
CONDORG_CLUSTER         I       -               GLIDEIN_ClusterId               Y       Y       @
CONDORG_SUBCLUSTER      I       -               GLIDEIN_ProcId                  Y       Y       @
CONDORG_SCHEDD          S       -               GLIDEIN_Schedd                  Y       Y       @
SEC_DEFAULT_ENCRYPTION  C       OPTIONAL        +                               N       N       -
SEC_DEFAULT_INTEGRITY   C       REQUIRED        +                               N       N       -
MAX_MASTER_LOG          I       1000000         +                               N       N       -
MAX_STARTD_LOG          I       10000000        +                               N       N       -

Glidein Variables

This section defines all the variables that the glideins explicity use. Please be aware that, apart from the below mentioned variable many other variables will be used by the HTCondor daemons, since glideins are HTCondor based; see the HTCondor manual for more details.

The variables can be divided based on their source:

Factory variables (config - attr tags)

This section presents variables that can be directly changed by a Glidein Factory administrator using attr tags in the Factory configuration XML using the following tags:

<attr name=“namevalue=“valtype=“type” .../>

Attr tags can be in both the Factory and Frontend configuration, available types are Int (I in HTCondor vars file), String (S) or Expr (C). String and Expr values are treated literally (no need to escape them) and Strings are quoted when passed to HTCondor.

More information on the XML format can be found in Glidein Factory configuration section.
If not specified in the XML, most of these variables have defaults set in condor vars files, which are used if the Glidein Factory administrator does not override them. These defaults are listed below.

Please also note that some of these variables may also be provided by the VO clients (e.g. Frontends). See the sections below.

Name

Type

Default Value

Description

GLIDEIN_Site String Entry name

Logical name of the Grid site where the glidein is running. This information is published both in the startd ClassAd and in the user job environment.

GLIDEIN_Hold Expr (Bool) True

HTCondor expression to use to specify when a user job in the glideins should be held. If any expression is true, the glidein is held. This is usually done to specify "bad" jobs, such as those that claim too much memory.

GLIDEIN_Entry_PREEMPT Expr (Bool) True

HTCondor expression to use to specify when a user job in the glideins should be preempted. If any expression is true, the glidein is preempted. This is usually done to specify custom preemption policies for user jobs.

GLIDEIN_PREEMPT Expr (Bool) True
GLIDEIN_Rank Expr (Int) 1

Used in calculating the HTCondor RANK

They are summed together, and the user job with the largest rank will run first.

GLIDEIN_Entry_Rank Expr (Int) 1
GLIDEIN_Max_Idle Int 1200
(20 mins)

Max amount of time a condor_startd will wait to be matched before giving up and terminating.

GLIDEIN_Max_Tail Int 400
(6 mins)

Max amount of time a condor_startd will wait after having already completed a job to be matched again. (i.e. the tail of a job).

GLIDEIN_Retire_Time Int 21600
(6 hours)

How long the condor_startd be running before no longer accepting jobs.
The random spread is used to improve the efficiencies, so the actual value used by HTCondor will be anywhere between GRT-GRTS and GRT.

NOTE: if GLIDEIN_Max_Walltime is specified, then this value is ignored and the retire time is calculated (see GLIDEIN_Max_Walltime below)

GLIDEIN_Retire_Time_Spread Int 7200
(2 hours)
GLIDEIN_Max_Walltime Int N/A

Max allowed time for the glidein.
This variable is optional. If you specify this variable, then HTCondor startup scripts will calculate the GLIDEIN_Retire_Time for the glidein as GLIDEIN_MAX_Walltime - GLIDEIN_Job_Max_Time. If GLIDEIN_Retire_Time is also specified, it will be ignored and only the calculated value is used. Note that if GLIDEIN_Job_Max_Time is specified but is greater than GLIDEIN_Max_Walltime or does not allow enough time for graceful shutdown, it will be reduced by the glidein. See Lifetime of a glidein for details.

GLIDEIN_Graceful_Shutdown Int 120

Once DAEMON_SHUTDOWN is reached and the glidein pilot enters the Retiring state, this amount passes to allow the startd and job to gracefully shutdown before forcefully terminating the glidein. See Lifetime of a glidein for details.

PREEMPT Expr (Bool) False Specifies whether preemption is allowed to occur on this glidein. N.B.: Starting from 3.4.3, glideins in the Factory queues are removed 12 hours after they hit the walltime. This is done with periodic remove expression, and happens even if PREEMPT is false.
PREEMPT_GRACE_TIME Int 10000000 This value affects the condor value "MaxJobRetirementtime" and it's an integer value representing the number of seconds a preempted job will be allowed to run before being evicted. This only affects behaviour if PREEMPT=True. After 12 hours, a glidein is removed from the factory queue anyway.
HOLD_GRACE_TIME Int 0 This value affects the condor value "MaxJobRetirementtime" and is an integer value representing the number of seconds a job that triggers WANT_HOLD will be allowed to run before being evicted. This only affects behaviour if GLIDEIN_HOLD, GLIDEIN_Entry_HOLD, GLIDECLIENT_HOLD, or GLIDECLIENT_Group_HOLD are specified and become true. By default, these "bad" jobs are immediately evicted.
GLIDEIN_Monitoring_Enabled Expr (Bool) True

Ability to control whether the pseduo-interactive monitoring slot is started on the worker node. Set to False if you do not want the monitoring slot started.

GLIDEIN_Resource_Slots String None

Special purpose resources added in separate slots or the main slot of the glidein. The separate slots will never be available for the regular jobs. This string is a semicolon separated list of comma separated resource descriptions. Each resource description contains the name (case insensitive) and optionally the number of resource instances (default is 1 (*)), the total memory reserved (default is 128MB times the number of resources), a type to control how the resource is handled (see below for more), and the disk reserved (default is auto, where HTCondor splits evenly). Possible type values are:

  • main - the resource is added to the main slot
  • extra - the resource is added to the main slot with virtual CPUs
  • partitionable - the resource has a new dedicated slot with virtual CPUs
  • static - the resource has many dedicated slots (one per instance) with a virtual CPU each

The default type is partitionable, unless there is only one resource instance, then is static. When adding resources to the main slots (type is main or extra): the memory and disk parameters are ignored; HTCondor splits automatically resources depending on the number of CPUs and having a partitionable main slot or not; if the number of resources is not equal (or an exact multiple) to the number of CPUs, then you must select partitionable slots for the Glideins (slots_layout="partitionable" in the config/submit section of the entry configuration or SLOTS_LAYOUT partitionable in the Frontend configuration) otherwise the startd may fail due to impossible configuration. When adding resources to the main slots, regular jobs (not using the resource) may use all memory and CPUs (including the virtual ones). Only separate slots (type is partitionable or static) reserve CPUs and memory. Check the HTCondor manual to learn more on how resources are split.

The parameters in a resource description can be listed or specified using their name: name, number, memory, type, disk (see the last example below).

Characters may be appended to a numerical value of memory to indicate percentage (%) or units. K or KB indicates KiB, $2^{10}$ numbers of bytes. M or MB indicates MiB, $2^{20}$ numbers of bytes. G or GB indicates GiB, $2^{30}$ numbers of bytes. T or TB indicates TiB, $2^{40}$ numbers of bytes. Disk space must be a fraction or percentage: 1/4, 25%. auto lets HTCondor do the splitting. Check the HTCondor manual for more about resource splitting.

Jobs submission must use request_RESOURCE=N with N>0 to use these slots or to use the resource in the main slot, e.g. request_ioslot=1

Examples:

  • value="ioslot,2,200,static" creates statically partitioned 2 ioslot resources with 100MB memory each.
  • value="ioslot" creates 1 ioslot resource with 128MB
  • value="ioslot,2,disk=1GB;monitor;GPUs,3,,main" creates (1) a partitionable slot with 2 ioslot resources, 256MB total memory and 1GB total disk (2) 1 monitor slot resource with 128MB memory (3) adds 3 GPUs to the main slot/slots
  • value="GPUs,type=main" discovers (*) how many GPUs are available and adds them to the main slot

(*) GPUs is a special resource name. If no number is specified, the glidein will invoke HTCondor's GPU discovery mechanism and get the number from there, which could be 0 if there are no GPUs. Your job will find in the HTCondor ClassAd also all the special attributes about the GPUs. If no number is specified and the auto-discovery fails 0 GPUs are assumed. Since you don't know the number of GPUs that will be added to the main slot, the slot_layout must be partitionable.

GLEXEC_BIN String None

If set, HTCondor will launch all user jobs via glexec, thus running the job under the appropriate local account. This is important both for GlideinWMS security and for accounting purposes.
A special value of "OSG" can be used to automatically find the locally installed gLExec on OSG worker nodes.

This variable is renamed to GLEXEC in the condor config.

GLEXEC_JOB Expr (Bool) False

If set to False, the condor_starter is run sharing the same UID as the user job. This has security implications.

NOTE: It is recommended to turn this on and have the condor_starter be protected from the user jobs.

GLIDEIN_Use_PGroups Expr (Bool) False

Should process group monitoring be enabled?

This is a HTCondor optimization parameter. Unfortunately, it negatively interferes with the batch systems used by the Grid sites, so it should not be turned on unless you have a very good reason to do so.

UPDATE_COLLECTOR_WITH_TCP Expr (Bool) True

If True, forces the glidein to use TCP updates.
The collector must be configured in the same way for this to work.

Also see the HTCondor documentation for implications and side effects.

WANT_UDP_COMMAND_SOCKET Expr (Bool) False

If True, enable the startd UDP command socket (HTCondor default).

Using the UDP command socket is a HTCondor optimization that makes working over firewalls and NATs very difficult. It is thus recommended you leave it disabled in the glideins.

Please note if you leave it disabled, that you must configure the schedd with
SCHEDD_SEND_VACATE_VIA_TCP = True
and the negotiator with
NEGOTIATOR_INFORM_STARTD = False
to have a functional system.

STARTD_SENDS_ALIVES Expr (Bool) True

If set to False, the schedd will be sending keepalives to the startd.

Setting this to True instructs the startd to send keepalives to the schedd. This improves the glidein behavior when running behind a firewall or a NAT.

Please note that the schedd must be configured in the same way for this to work.

SEC_DEFAULT_INTEGRITY Expr REQUIRED

Security related settings. Please notice that the glideins always require GSI authentication.

For more details see the configuration page or the HTCondor manual.

SEC_DEFAULT_ENCRYPTION Expr OPTIONAL
USE_MATCH_AUTH Expr (Bool) False

Another security setting.

If set to True, the schedd and the startd will use a low overhead protocol. See the configuration page or the HTCondor manual.

MAX_MASTER_LOG Int 1M

What is the maximum size the logs should grow.

Setting them too low will made debugging difficult.
Setting them too high may fill up the disk in anomalous situations, both on the work nodes and on the glidein Factory.

MAX_STARTD_LOG Int 10M
MAX_STARTER_LOG Int 10M
USE_CCB Expr (Bool) False

If set to True, it will enable HTCondor Connection Brokering (CCB), which is needed if the glideins are behind a firewall or a NAT.

USE_SHARED_PORT Expr (Bool) False

If set to True, it will enable the shared port daemon, which will reduce the number of connections between the glidein and the collectors.

GLIDEIN_CPUS String 1

Number of CPUs glidein should use. GLIDEIN_CPUS is used to set NUM_CPUS for the HTCondor started by the glidein. Use "slot" (or "auto") to let glidein determine this from the job slot assigned to the glidein, use "node" to let glidein determine this from the hardware of the worker node (e.g. if you are sure that only your job is running on the node), use a positive integer to set GLIDEIN_CPUS (and NUM_CPUS) to assume that value. Most of the time it is an integer, but the type is string to allow also the keywords. In case of static partitioning, glidein will create GLIDEIN_CPUS number of slots. In case of dynamic partitioning, the slots will be created automatically based on the CPUs required by the job and GLIDEIN_CPUS is the sum of the slots made available. Refer to HTCondor manual for info on NUM_CPUS

GLIDEIN_ESTIMATED_CPUS Int 1

Number of CPUs glidein is expected to receive. GLIDEIN_ESTIMATED_CPUS is used only when GLIDEIN_CPUS is set to some auto-discovery value (auto/slot, node). With cores auto-discovery the Frontend assumes the entry will provide GLIDEIN_ESTIMATED_CPUS cores, 1 by default. The Frontend will work even if the estimate is not correct, anyway: it will trigger requests for the entry only if the CPUs requested by the jobs are no more than GLIDEIN_ESTIMATED_CPUS and, it will adjust the number of Glideins requested assuming GLIDEIN_ESTIMATED_CPUS cores from each Glidein in that entry. GLIDEIN_ESTIMATED_CPUS is not used by the Glidein. GLIDEIN_ESTIMATED_CPUS should be set only when GLIDEIN_CPUS is set to some auto-discovery value (auto/slot, node). It is a configuration error to set GLIDEIN_ESTIMATED_CPUS otherwise.

GLIDEIN_NODES Int 1

Number of NODEs a Glidein is expected to be replicated on. On some entries, especially HPC resources, each Glidein submission triggers the startup of the Glidein on multiple nodes. The Frontend assumes that for each requested Glidein the entry will start it on GLIDEIN_NODES nodes, 1 by default. The Frontend will work even if the estimate is not correct, anyway: it will trigger requests for the entry only if the CPUs requested by the jobs are no more than GLIDEIN_ESTIMATED_CPUS and, it will adjust the number of Glideins requested assuming GLIDEIN_NODES Glideins, each with GLIDEIN_CPUS/GLIDEIN_ESTIMATED_CPUS cores for each request to that entry. GLIDEIN_NODES is not used by the Glidein.

GLIDEIN_MaxMemMBs Int None

Amount of memory glidein should use. If set, GLIDEIN_MaxMemMBs is used to set total MEMORY used for the HTCondor Startd and jobs started by the glidein. If GLIDEIN_MaxMemMBs is not set and GLIDEIN_MaxMemMBs_Estimate is TRUE, GLIDEIN_MaxMemMBs is calculated based on the memory available on the worker node. If GLIDEIN_MaxMemMBs is not set and GLIDEIN_MaxMemMBs_Estimate is not TRUE (False or unset), the glidein lets HTCondor decide the amount of memory. Refer to HTCondor manual for info on MEMORY

GLIDEIN_MaxMemMBs_Estimate Expr (Bool) False

Used in conjunction with GLIDEIN_MaxMemMBs. It is ignored if GLIDEIN_MaxMemMBs is set to any value. See GLIDEIN_MaxMemMBs for the description.

GLIDEIN_Factory_Report_Failed String ALIVEONLY

This attribute regulates advertising of validation failures to the Factory collector.

  • NEVER - The glidein will just fail, and do nothing.
  • ALIVEONLY - The glidein will advertise a Master ClassAd in Drained state and Retiring activity for the lifetime of the glidein script.
  • ALWAYS - Similar to ALIVEONLY, but will also send a final ClassAd in Drained state and Killing activity just before dying.

When advertised, the classad is flagged GLIDEIN_Failed=True, the error is recorded in the GLIDEIN_FAILURE_REASON and GLIDEIN_EXIT_CODE attributes, and the failing script is recorded in the GLIDEIN_LAST_SCRIPT attribute.

Factory config xml - configuration

The second set of variables comes from values the Glidein Factory administrator defined to make the Factory to work. They are generated based on xml tags in the Factory configuration (most in the entry tag). They cannot be changed by an administrator in any other way.

Name Type Source Description
GLIDEIN_Factory String <glidein factory_name="value">

Logical name of the Glidein Factory machine (like “osg1”).

GLIDEIN_Name String <glidein glidein_name="value">

Identification name of the Glidein Factory instance (like “v1_0”).

GLIDEIN_Entry_Name String ...<entries><entry name=”value”>

Identification name of the entry point (like “ucsd5”).

GLIDEIN_GridType String ...<entries><entry gridtype=“value”>

Type of Grid resource (like “condor”).

GLIDEIN_Gatekeeper String ...<entries><entry gatekeeper=“value”>

URI of the Grid gatekeeper (like “osg1.ucsd.edu/jobmanager-pbs”)

GLIDEIN_GlobusRSL String ...<entries><entry rsl=“value”>

Optional RSL string (like "(condor_submit=('+ProdSlot' 'TRUE'))")

PROXY_URL String ...<entries><entry proxy_url=“value”>

Optional URL of the site Web proxy.

A special value “OSG” can be used to automatically discover the local Web proxy on OSG worker nodes.

This variable is exported as GLIDEIN_Proxy_URL to the use job environment.

DEBUG_MODE String ...<entries><entry verbosity=“value”>

This setting can be either:

  • “std” - Default mode, where all interesting debug information is reported back to the Glidein Factory and the glidein will wait 20 minutes on a worker node that failed validation to minimize the black hole effect.

  • “nodebug” - Disable most diagnostic messages. This can be useful for very stable setups. The glidein still waits 20 minutes on a worker node that failed validation to minimize the black hole effect.

  • Fast – All debugging is enabled and the glidein waits only 2 minutes on a worker node that failed validation. This mode is useful when debugging a misbehaving Grid site.

Factory or Frontend variables (config - attr tags)

This set of values may come from the Glidein Factory or Frontend configuration. While a client can set any number of variables, the ones described below are the most often used. The value set in the Factory can be overridden by the Frontend, unless set constant in the Factory. Some of the values below must be a parameter (i.e. have parameter="True" attr specification). This is because they are used in the early phases of the Glidein execution. NOTE that as of 3.4.6, in the Factory, also publish=True and const=False should be set, otherwise parameter=True will be ignored.

Name Type Description
GLIDEIN_DEBUG_OUTPUT Str

If True enables debug output from the Glidein (more verbose messages). The messages printed are affected also by GLIDEIN_DEBUG_OPTIONS. If not empty is considered True.

GLIDEIN_DEBUG_OPTIONS Str

Comma separated list of keywords affecting Glidein debug. Possible options are:
userjob - prints debug informations on the user job stderr (at the beginning, when invoked with the Singularity wrapper)

GLIDEIN_MULTIGLIDEIN Int

Forks multiple glideins, each in its own subdirectory, all w/ the same parameters (similar to multinode submission, but is all on one node). Unlsess GLIDEIN_MULTIGLIDEIN_LAUNCHALL is set, then a single invocation is used, expecting the system to handle the spawning. Must be a parameter.

GLIDEIN_MULTIGLIDEIN_LAUNCHER String

Used only if GLIDEIN_MULTIGLIDEIN is set and > 1, and GLIDEIN_MULTIGLIDEIN_LAUNCHALL is not set. Defines a launcher program that is added at the beginning of the command line when each multiple Glidein instance is started. NOTE: parameter and job_publish should be both True for this attribute. Must be a parameter.

GLIDEIN_MULTIGLIDEIN_LAUNCHALL String

Used only if GLIDEIN_MULTIGLIDEIN is set and > 1. When this is set, a command line with this string followed by the glidein invocation is used to start all Glideins. NOTE: parameter and job_publish should be both True for this attribute. Must be a parameter.

Frontend Client Variables

The third set of values comes from the Glidein Frontend clients. While a client can set any number of variables, the ones described below are the most often used.

Name Type Description
GLIDEIN_Client String

Identification name of the VO Frontend request (like “ucsd5@v1_0@osg1@cms4”).

GLIDEIN_Collector Expr (List)

List of Collector URIs used by the VO Condor pool (like “cc.cms.edu:9620,cc.cms.edu:9621”).

One of the URIs in the list will be selected and used as HEAD_NODE in the condor_config.

GLIDECLIENT_Hold Expr (Bool)

Condor expression to use to specify when a user job in the glideins should be held. If any expression is true, the glidein is held. This is usually done to specify "bad" jobs, such as those that claim too much memory.

GLIDECLIENT_Group_Hold Expr (Bool)
GLIDECLIENT_PREEMPT Expr (Bool)

Condor expression to use to specify when a user job in the glideins should be preempted. If any expression is true, the glidein is preempted. This is usually done to specify custom preemption policies.

GLIDECLIENT_Group_PREEMPT Exp (Bool)
GLIDECLIENT_Rank Expr (Int)

Used in calculating the Condor RANK

They are summed together, and the user job with the largest rank will run first.

GLIDECLIENT_Group_Rank Expr (Int)
GLIDEIN_Ignore_X509_Duration Expr (Bool)

True by default. Setting this to false will shorten the Glidein lifetime to be shorter than the Proxy lifetime. Use it if you expect the resource will not extend reliably the proxy lifetime. The Glidein lifetime will always be shorten if the Proxy lifetime is less than 900 s.

GLIDEIN_Job_Max_Time Int

Max allowed time for the job to end.
This is used by the condor_startd. Once the GLIDEIN_Retire_Time expires, the glidein will allow the job to run for this amount of time before activating DAEMON_SHUTDOWN and killing the job. Note that, if GLIDEIN_Max_Walltime is specified, this variable could be reduced by the glidein in order to fit within Glidein_Max_Walltime. See Lifetime of a glidein for details.

GLIDEIN_Expose_Grid_Env Expr (Bool)

If False, the user job environment will contain only glidein Factory provided variables.

If True, the user job environment will also contain the environment variables defined at glidein startup.

See JOB_INHERITS_STARTER_ENVIRONMENT documentation for more details.

GLIDEIN_Expose_X509 Expr (Bool)

By default, the glidein will unset the variable X509_USER_PROXY for security reasons to prevent the user jobs from accessing the pilot proxy. Setting this to true will override this behavior and keep the X509_USER_PROXY in the environment.

SLOTS_LAYOUT String Defines how multi-core glideins should split their resources.

There are only two legal values: The default behavior is defined by the Factory, but it has historically been set to "fixed".

Note: This variable MUST NOT be passed as a parameter, or the glideins will fail!
FORCE_PARTITIONABLE String (Bool) By default, single core glideins will never be configured as partitionable, independently of the value of SLOTS_LAYOUT.

If partitionable slots are desired also for single-core glideins, set this variable to "True".
GLIDEIN_Report_Failed String

This attribute regulates advertising of validation failures to the user collector.

  • NEVER - The glidein will just fail, and do nothing. (default)
  • ALIVEONLY - The glidein will advertise a Machine ClassAd in Drained state and Retiring activity for the lifetime of the glidein script.
  • ALWAYS - Similar to ALIVEONLY, but will also send a final ClassAd in Drained state and Killing activity just before dying.

When advertised, the classad is flagged GLIDEIN_Failed=True, the error is recorded in the GLIDEIN_FAILURE_REASON and GLIDEIN_EXIT_CODE attributes, and the failing script is recorded in the GLIDEIN_LAST_SCRIPT attribute.

GLIDEIN_CLAIM_WORKLIFE Int

CLAIM_WORKLIFE for non-dynamic slots. Defaults to -1 i.e. HTCondor will treat this as an infinite claim worklife and schedd will hold claim to the slot until jobs are preempted or user runs out of jobs.

GLIDEIN_CLAIM_WORKLIFE_DYNAMIC Int

CLAIM_WORKLIFE for dynamically partitionable slots. Defaults to 3600. This controls how frequently the dynamically partionable slots will coalesce.

GLIDEIN_Custom_Start Expr (Bool)

Defaults to true. An expression that gets appended to the start expression of the startd daemons that the pilots start on the worker node. Any frontend validation script can modify this and customize the set of user jobs that are matching. This is useful in order to have only a certain set of user jobs (e.g.: only montecarlo jobs) matching pilots at particular resources (e.g.: sites with slow network/disk)

Cloud VM specific Variables

The following variables are only applicable to Cloud VMs.

These variables can be either configured by Factory or the Frontend

Name

Type

Default Value

Description

VM_MAX_LIFETIME Int 172800
(48 hours)

Max lifetime of the VM. When this timer is reached the glideinwms-pilot service will terminate and glidein process, shutdown the glideinwms-pilot service and issue a VM shutdown

VM_DISABLE_SHUTDOWN Expr (Bool) False

Disables VM from automatically shutting down after the glideinwms-pilot service has exited.
NOTE: This variable is for debugging purposes only and should not be set to True in the Production environment. Disabling auto shutdown could result in accumulating financial charges in case of paid clouds.

Dynamically generated variables

The following variables are being dynamically generated and/or modified by GlideinWMS processes. The GlideinWMS administrators cannot directly change them.

The first set of variables comes from the Glidein Factory.

Name Type Description
GLIDEIN_Signature String

These variables contain the SHA1 signature of the signature files.

These signatures are used as a base to ensure the integrity of all the data downloaded in the glidein startup scripts, but they also provide a fingerprint of the configuration used by the glidein.

These variables are published both in the glidein ClassAd and in the user job environenmt.

GLIDEIN_Entry_Signature String
CONDORG_SCHEDD String

The schedd used by the Glidein Factory to submit the glidein.

This variables is exported a GLIDEIN_Schedd both in the glidein ClassAd and to the user job environment.

CONDORG_CLUSTER Int

The cluster and process id assigned by the Glidein Factory schedd to this glidein.

These variables are exported as GLIDEIN_ClusterId and GLIDEIN_ProcId both in the glidein ClassAd and to the user job environment.

CONDORG_SUBCLUSTER Int

Directory Path Variables

The next set contains the location of files and/or directories downloaded or created by the glidein. Most of them are located under the working directory specified by

<entry work_dir=“value”>
Name Description
TMP_DIR

Path to the directory that admin-provided scripts and user jobs can use for storing temporary data.

This variable is exported as GLIDEIN_Tmp_Dir both to the glidein ClassAd and to the user job environment.

GLIDEIN_LOCAL_TMP_DIR

Path to the directory on the local file system for storing temporary data.
Note: Only use for things that absolutely need the guarantee of residing in the local file system. Prefer TMP_DIR above for everything else.

This variable is exported to the user job environment.

CONDOR_VARS_FILE

File path to the condor vars files.

Admin-provided scripts may want to add entries to these files.

CONDOR_VARS_ENTRY_FILE
ADD_CONFIG_LINE_SOURCE

File path to the script containing the add_config_line and add_confir_vars line functions.

X509_USER_PROXY

File path to the glidein proxy file.

X509_CONDORMAP

File path to the Condor mapfile used by the glidein.

X509_CERT_DIR

Path to the directory containing the trusted CAs' public keys and RSLs.

CONDOR_DIR

Directory where the glidein Condor binary distribution have been installed.

WRAPPER_LIST

File path to the list of wrapper scripts used by the glidein.

GLIDEIN_WRAPPER_EXEC

This is the executable that glideins will run (ie what to put after "exec" in the condor job wrapper). By default, glideins will perform exec "\$@" to run the pilot. For other modes of execution, you may need different arguments. For example, to run a program under parrot, you may need exec "$GLIDEIN_PARROT/parrot_run" -t "$parrot_tmp" "$@" .
Note, the syntax of this command is very sensitive This needs to be double-escaped in order to function correctly. For the two examples listed above, you would need:

  • "\\\"\\\$@\\\""
  • "\\\"\\\$GLIDEIN_PARROT/parrot_run\\\" -t \\\"$parrot_tmp\\\" \\\"\\\$@\\\""

Machine Job Features variables (dynamic)

This set of variables contains various variables generated by the glidein startup and periodic scripts, that are related to the Machine Job Features. The content of these variables can only be a number, or an a string that does not contain double quote or backslash characters. Otherwise, the variable content will be set to the string "WrongFormat". For convenience the description of those variable is replicated here in this page.

Name Type Description
MJF_MACHINE_TOTAL_CPU Int

Number of processors which may be allocated to jobs. Typically the number of processors seen by the operating system on one worker node (that is the number of \processor :" lines in /proc/cpuinfo on Linux), but potentially set to more or less than this for performance reasons.

MJF_MACHINE_HS06 Int

Total HS06 rating of the full machine in its current setup. HS06 is measured following the HEPiX recommendations, with HS06 benchmarks run in parallel, one for each processor which may be allocated to jobs.

MJF_MACHINE_SHUTDOWNTIME Int

Shutdown time for the machine as a UNIX time stamp in seconds. The value is dynamic and optional. If the file is missing, no shutdown is foreseen.

MJF_MACHINE_GRACE_SECS Int

If the resource provider announces a shutdown time to the jobs on this host, that time will not be less than grace secs seconds after the moment the shutdown time is set. This allows jobs to begin packages of work knowing that there will be sufficient time for them to be completed even if a shutdown time is announced. This value is required if a shutdown time will be set or changed which will affect any jobs which have already started on this host.

MJF_JOB_ALLOCATED_CPU Int

Number of processors allocated to the current job.

MJF_JOB_HS06_JOB Int

Total HS06 rating for the processors allocated to this job. The job's share is calculated by the resource provider from per-processor HS06 measurements made for the machine.

MJF_JOB_SHUTDOWNTIME_JOB Int

Dynamic value. Shutdown time as a UNIX time stamp in seconds. If the file is missing no job shutdown is foreseen. The job needs to have finished all of its processing when the shutdown time has arrived.

MJF_JOB_GRACE_SECS_JOB Int

If the resource provider announces a shutdowntime job to the job, it will not be less than grace secs job seconds after the moment the shutdown time is set. This allows jobs to begin packages of work knowing that there will be sufficient time for them to be completed even if a shutdown time is announced. This value is static and required if a shutdown time will be set or changed after the job has started.

MJF_JOB_JOBSTART_SECS Int

UNIX time stamp in seconds of the time when the job started on the worker node. For a pilot job scenario, this is when the batch system started the pilot job, not when the user payload started to run.

MJF_JOB_ID Int

A string of printable non-whitespace ASCII characters used by the resource provider to identify the job at the site. In batch environments, this should simply be the job ID. In virtualized environments, job id will typically contain the UUID of the VM.

MJF_JOB_WALL_LIMIT_SECS Int

Elapsed time limit in seconds, starting from jobstart secs. This is not scaled up for multiprocessor jobs.

MJF_JOB_CPU_LIMIT_SECS Int

CPU time limit in seconds. For multiprocessor jobs this is the total for all processes started by the job

MJF_JOB_MAX_RSS_BYTES Int

Resident memory usage limit, if any, in bytes for all processes started by this job.

MJF_JOB_MAX_SWAP_BYTES Int

Swap limit, if any, in bytes for all processes started by this job.

MJF_JOB_SCRATCH_LIMIT_BYTES Int

Scratch space limit if any. If no quotas are used on a shared system, this corresponds to the full scratch space available to all jobs which run on the host. User jobs from EGI-registered VOs expect the \max size of scratch space used by jobs" value on their VO ID Card to be available to each job in the worst case. If there is a recognised procedure for informing the job of the location of the scratch space (eg EGI's $TMPDIR policy), then this value refers to that space.

Singularity Variables (dynamic)

These variables can be set in the Factory attributes, set in the Frontend attributes, set in the Job (submit file), or cannot be set and is dynamically generated by the startup scripts (Dynamic). A value can be also be a combination of the above: the Factory value can be overridden by the Frontend one (unless set as constant), and the startup scripts can modify it and have the ultimate saying.

The resulting invocation of Singularity is something like the text below. Where $singularity_binds is built using GLIDEIN_SINGULARITY_BINDPATH_DEFAULT, GLIDEIN_SINGULARITY_BINDPATH, and GWMS_SINGULARITY_BIND_CVMFS, and the $singularity_image is one coming from SINGULARITY_IMAGES_DICT or specified in the job. The debug options are used if GLIDEIN_DEBUG_OUTPUT is set.

 singularity [-d -vvv] exec --home "$PWD":/srv --pwd /srv --ipc --pid --contain \
    $GLIDEIN_SINGULARITY_OPTS --bind "$singularity_binds"} \
    "$singularity_image" "${@}"

Name

Type

Default Value

Where

Description

GLIDEIN_SINGULARITY_REQUIRE String (Enum) Factory

An entry can control the use of Singularity by setting GLIDEIN_SINGULARITY_REQUIRE to NEVER (Singularity is not supported), OPTIONAL or PREFERRED (capable of Singularity but it is not enforced), REQUIRED (jobs must run with Singularity) or REQUIRED_GWMS (jobs must run with Singularity and use the GWMS wrapper scripts). This last option is the only one that really enforces Singularity, but is not compatible with VOs that currently self-manage Singularity with custom scripts, like OSG and CMS.

GLIDEIN_Singularity_Use String (Enum) DISABLE_GWMS Frontend

Determines whether or not Frontend/Group want to mandate the use of Singularity and is used both for provisioning and once the job is on the site. Possible values are: DISABLE_GWMS (disable GWMS Singularity machinery, not to interfere with independent VO setups - Default), or NEVER (do not use Singularity), or OPTIONAL or PREFERRED (use Singularity if the site is configured with it), or REQUIRED (always require Singularity). If REQUIRED, PREFERRED or OPTIONAL are used, the default wrapper script that we provide should be used (or a similar one). In other words a line like <file absfname="/var/lib/gwms-frontend/web-base/frontend/default_singularity_wrapper.sh" wrapper="True"/> should be added in the <files> section of the general or group parts of the Frontend configuration.

GWMS_SINGULARITY_STATUS String (Enum) Dynamic

Is evaluated using GLIDEIN_Singularity_Use and GLIDEIN_SINGULARITY_REQUIRE and determines whether and how Singularity will be used. Possible values are: DISABLE (do not use the GWMS Singularity mechanisms), FAIL (do not provision on this entry, do not match with these Glideins or fail the Glidein), NEVER (do not use Singularity), PREFERRED (use Singularity but allow to fall-back to no Singularity if something goes wrong, e.g. no image is found), REQUIRED (use Singularity and fail the Glidein if it cannot run with Singularity).
See the Factory configuration document for a table of how Singularity is calculated using GLIDEIN_Singularity_Use and GLIDEIN_SINGULARITY_REQUIRE.

HAS_SINGULARITY Expr (Bool) Dynamic

Tells the Glidein wheter Singularity is available. This is not set if GWMS_SINGULARITY_STATUS is DISABLE, it is set to 1 (true) if GWMS_SINGULARITY_STATUS is PREFERRED or REQUIRED, and it is set to 0 (false) otherwise. It can change to 0 if GWMS_SINGULARITY_STATUS is PREFERRED but the Singularity tests fail.

SINGULARITY_IMAGES_DICT String Frontend Factory

String-encoded comma separated key-value pair dictionary listing Singularity images. Keys cannot contain commas or colons. Values cannot contain commas. The keys are identifiers for a platform and the values are paths (or URLs) for the corresponding images, expressed in any format supported by Singularity. REQUIRED_OS is used to select the platform: the first platform in the list that matches a key will be used. You can use any string to identify the images as lons as you are consistent with REQUIRED_OS, but we recommend to use standard names like: rhel7 (for RHEL7 or SL7), rhel6 (fpr RHEL6 or SL6). The default set in generic_pre_singularity_setup.sh is "rhel6:/cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-el6:latest,rhel7:/cvmfs/singularity.opensciencegrid.org/opensciencegrid/osgvo-el7:latest" or by running a pre_singularity setup script. SINGULARITY_IMAGE_DEFAULT6, SINGULARITY_IMAGE_DEFAULT7, and SINGULARITY_IMAGE_DEFAULT are deprecated and will be ignored in the future. To ease the transition, these are currently added to SINGULARITY_IMAGES_DICT with keys rhel6, rhel7 and default, respectively.

REQUIRED_OS String any Frontend Job

Comma separated list of platforms accepted by the job. 'any' is a wildcard, means that any platform is OK. GWMS will intersect this list with GLIDEIN_REQUIRED_OS and use the intersection to select the Singularity image. The first value of the intersection for which there is a key in SINGULARITY_IMAGES_DICT will determin the image (value with that key). 'any' means that any Singularity image is OK.

GLIDEIN_REQUIRED_OS String any Factory

Comma separated list of platforms accepted by the entry. See REQUIRED_OS.

SINGULARITY_IMAGE_RESTRICTIONS String (Enum) any Factory Frontend

Comma separated list of restrictions on the Singularity images. Possible values are: cvmfs (the image must be on cvmfs).

GLIDEIN_SINGULARITY_BINDPATH_DEFAULT String Factory

This is used to set Singularity bind mounts. It is a comma separated list of src[:dst[:opt]] elements. See GLIDEIN_SINGULARITY_BINDPATH

GLIDEIN_SINGULARITY_BINDPATH String Frontend

This is used to set Singularity bind mounts. It is a comma separated list of src[:dst[:opt]] elements, as you would express in the --bind option in Singularity. 'src' is the path on the system, 'dst' is the path in the Singularity image, 'opt' is a mount option (e.g. 'ro', for read-only mounts). See the Singularity for a description of the format and the supported options. Commas are not allowed in both the paths, and colons are not allowed in the 'src' path. GWMS combines in order GLIDEIN_SINGULARITY_BINDPATH, GLIDEIN_SINGULARITY_BINDPATH_DEFAULT, and a default list. The default list includes "/hadoop,/hdfs,/lizard,/mnt/hadoop,/mnt/hdfs", and "/cvmfs" (if GWMS_SINGULARITY_BIND_CVMFS is true) and some GPUs libraries (if GPUs are detected). The combined list is validated removing paths not existing (outside Singularity) and is passed as --bind option to singularity. Singularity prepends to the mounts list also the environment variable $SINGULARITY_BINDPATH (this could be set by the Site or by setup scripts on the worker nodes). Singularity will skip mounts where the destination path does not exist in the Singularity image selected. The suggestion is to set GLIDEIN_SINGULARITY_BINDPATH_DEFAULT in the Factory and GLIDEIN_SINGULARITY_BINDPATH in the Frontend.

GWMS_SINGULARITY_BIND_CVMFS Expr (Bool) 1 Frontend Factory

/cvmfs is mounted in Singularity if this is 1 (true)

GLIDEIN_SINGULARITY_OPTS String Frontend Factory

Extra options that will be added to the singularity invocation

Glidein Script Variables (dynamic)

The last set contains various variables generated by the glidein startup scripts.

Name Type Description
X509_GRIDMAP_DNS String

List of DNs trusted by the glidein.

X509_EXPIRE Expr (time_t)

When is the proxy expected to expire.

GLEXEC_STARTER Expr (Bool)

If gLExec is used and this is set to True, the condor_starter will be run sharing the same UID as the user job.

ALTERNATIVE_SHELL String

If gLExec is used, this variable points to a trusted copy of a shell.

GLEXEC_USER_DIR String

If gLExec is used, this variable points to the working directory under which all user jobs will be started.

SiteWMS_WN_Draining Expr (Bool)

The variable controls wether or not the glidein should accept new jobs. As part of the WLCG Machine / Job Features Task Force Site admins have the possibility to put worker nodes to "drain mode". In particular if the JOBFEATURES (or MACHINEFEATURES) environment variable is set and it points to a local directory containing a file named shutdowntime_job (shutdowntime), or $JOBFEATURES/shutdowntime_job ($MACHINEFEATURES/shutdowntime) is a valid URL (it is possible to wget it), then the glidein will stop accepting jobs. (see here). The SiteWMS_WN_Draining variable is periodically updated by means of the STARTD_CRON condor feature. See Draining glideins for details.

SiteWMS_WN_Preempt Expr (Bool)

The variable controls wether or not the glidein should preempt jobs. This is still part of the WLCG Machine / Job Features Task Force (see above). The variable will become true if the shutdown time value contained in the file pointed by the JOBFEATURES (or MACHINEFEATURES) variable contains a timestamp (unix epoch time) that is less than 30 minutes in the future. If so PREEMPT_GRACE_TIME will be set to 20 minutes and the job will be preempted after that time if it does not exit. The SiteWMS_WN_Preempt variable is periodically updated by means of the STARTD_CRON condor feature. See Draining glideins for details.

Set and discover available CPUs

For this explanation the term CPUs refers to the number of cores, provided by one or more CPUs. It is whatever is detected or provided and will set the level of parallelism (NUM_CPUS) used by the HTCondor startd started by the Glidein.

Jobs express the number of CPUs they require via RequestCpus.

You can force HTCondor to assume it has a certain number of CPUs by setting GLIDEIN_CPUS to a number or let the Glidein discover the number of physical CPUs in the worker node (node) or the number of CPUs received from the job manager (slot, auto).

If GLIDEIN_CPUS is set to auto (or slot, node). When the Frontend does provisioning, it triggers requests to resources where RequestCpus<=GLIDEIN_ESTIMATED_CPUS (1 if GLIDEIN_ESTIMATED_CPUS is not set) and it is dimensioning the request expecting to receive CPUs = GLIDEIN_ESTIMATED_CPUS * number_of_glidiens.
When Glideins actually start at the site, they advertise the CPUs they actually discover, ignoring what is set in GLIDEIN_ESTIMATED_CPUS and will match jobs with RequestCpus<=CPUs discovered. This is different from setting GLIDEIN_CPUS, that artificially tells HTCondor in the Glidein that it has those CPUs, over- or under-provisioning if the number of CPUs discovered is different.

So if you have a site with a variable number of CPUs you can set GLIDEIN_ESTIMATED_CPUS to the min number of CPUs you will be receiving (for ease call it MinCpus). All jobs triggering Glideins on that site will be able to run on them. Multiple jobs will run on the same Glidein if it receives more CPUs. If you need to run jobs w/ RequestCpus>MinCpus, you can set GLIDEIN_ESTIMATED_CPUS to a bigger number but be aware that some Glideins (the ones that received less than GLIDEIN_ESTIMATED_CPUS CPUs) will not match jobs that triggered their submission (those Glideins will probably be wasted).

Another attribute changing the number of CPUs is GLIDEIN_Resource_Slots. When you add resource slots of type extra, partitionable or static, these increase the number of CPUs set in HTCondor, independently from whether the initial number was set with GLIDEIN_CPUS or discovered. These virtual CPUs are though not considered in provisioning, are frequently reserved for jobs matching the special resources (partitionable, static) and are available also for regular jobs only when the slot type extra is used.

For more information, see the descriptions of GLIDEIN_CPUS, GLIDEIN_ESTIMATED_CPUS, and GLIDEIN_Resource_Slots above. And NUM_CPUS in the HTCondor manual.

Lifetime of a Glidein

Retire time plot

All the various variables in the glidein configuration can be confusing. The above diagram illustrates the lifetime of a glidein pilot that has a long-running job.

  • Green: For the first GLIDEIN_Retire_Time seconds (modified by a random spread GLIDEIN_Retire_Spread to smooth out glideins all ending simultaneously), jobs can start on the glidein pilot.
  • Yellow: During the yellow period, START will evauluate to FALSE, so no new jobs will start. However, already running jobs will continue to run for GLIDEIN_Job_Max_Time. (Note that the glideins will end during this period if the job ends. They will not idle afterwards since no new jobs can start anyway).
  • Once this period is done, the DAEMON_SHUTDOWN variable will be true.
  • Orange: There can be a delay of up to UPDATE_INTERVAL (usually about 5 minutes) between when DAEMON_SHUTDOWN becomes true and when it is actually updated in the collector. This is because the collector only reevaluates this expression periodically.
  • Red: Once DAEMON_SHUTDOWN is true, condor gives a short grace period of GLIDEIN_Graceful_Shutdown before forcefully terminating everything and shutting down. All of these periods are totalled and calculated to fit within GLIDEIN_MAX_WALLTIME if specified.

Note that if a job ends early in the green period, a new job will start. If a job ends after this period, then the glidein will shut down and end early. This can be seen in the example below with two jobs below:

Idle time plot

  • Orange: A glidein first starts up.
  • Yellow: After starting up, the glidein will wait for Glidein_Max_Idle for its initial matching. If it is Idle for longer than this, it will assume no jobs are available and will shutdown.
  • Green: A job runs (subject to the maximum limits in the previous diagram).
  • Yellow: Once a job runs, it will wait for Glidein_Max_Tail for another job.
  • Green: Another matching job runs.
  • Yellow: Once the job finishes, it will wait for Glidein_Max_Tail for another job. If cannot find one, it will shutdown.
Glidein_Max_Tail counts the idle time after the previous job ends, as specified above. N.B.: 12 hours after the walltime, a glidein is removed from the factory queue no matter what is using a periodic remove expression. This is done to prevent staleness of glideins which are the glideins that appears running in the factory condor queue indefinitely, but are no longer in the CE queue.