A hook is an external program or script invoked by HTCondor.
Job hooks that fetch work allow sites to write their own programs or scripts, and allow HTCondor to invoke these hooks at the right moments to accomplish the desired outcome. This eliminates the expense of the matchmaking and scheduling provided by the condor_schedd and the condor_negotiator, although at the price of the flexibility they offer. Therefore, job hooks that fetch work allow HTCondor to more easily and directly interface with external scheduling systems.
Hooks may also behave as a Job Router.
The Daemon ClassAd hooks permit the condor_startd and the condor_schedd daemons to execute hooks once or on a periodic basis.
Note that standard universe jobs execute different condor_starter and condor_shadow daemons that do not implement any hook mechanisms.
In the past, HTCondor has always sent work to the execute machines by pushing jobs to the condor_startd daemon, either from the condor_schedd daemon or via condor_cod. Beginning with the HTCondor version 7.1.0, the condor_startd daemon now has the ability to pull work by fetching jobs via a system of plug-ins or hooks. Any site can configure a set of hooks to fetch work, completely outside of the usual HTCondor matchmaking system.
A projected use of the hook mechanism implements what might be termed a glide-in factory, especially where the factory is behind a firewall. Without using the hook mechanism to fetch work, a glide-in condor_startd daemon behind a firewall depends on CCB to help it listen and eventually receive work pushed from elsewhere. With the hook mechanism, a glide-in condor_startd daemon behind a firewall uses the hook to pull work. The hook needs only an outbound network connection to complete its task, thereby being able to operate from behind the firewall, without the intervention of CCB.
Periodically, each execution slot managed by a condor_startd will invoke a hook to see if there is any work that can be fetched. Whenever this hook returns a valid job, the condor_startd will evaluate the current state of the slot and decide if it should start executing the fetched work. If the slot is unclaimed and the Start expression evaluates to True, a new claim will be created for the fetched job. If the slot is claimed, the condor_startd will evaluate the Rank expression relative to the fetched job, compare it to the value of Rank for the currently running job, and decide if the existing job should be preempted due to the fetched job having a higher rank. If the slot is unavailable for whatever reason, the condor_startd will refuse the fetched job and ignore it. Either way, once the condor_startd decides what it should do with the fetched job, it will invoke another hook to reply to the attempt to fetch work, so that the external system knows what happened to that work unit.
If the job is accepted, a claim is created for it and the slot moves into the Claimed state. As soon as this happens, the condor_startd will spawn a condor_starter to manage the execution of the job. At this point, from the perspective of the condor_startd, this claim is just like any other. The usual policy expressions are evaluated, and if the job needs to be suspended or evicted, it will be. If a higher-ranked job being managed by a condor_schedd is matched with the slot, that job will preempt the fetched work.
The condor_starter itself can optionally invoke additional hooks to help manage the execution of the specific job. There are hooks to prepare the execution environment for the job, periodically update information about the job as it runs, notify when the job exits, and to take special actions when the job is being evicted.
Assuming there are no interruptions, the job completes, and the condor_starter exits, the condor_startd will invoke the hook to fetch work again. If another job is available, the existing claim will be reused and a new condor_starter is spawned. If the hook returns that there is no more work to perform, the claim will be evicted, and the slot will return to the Owner state.
There are a handful of hooks invoked by HTCondor related to fetching work, some of which are called by the condor_startd and others by the condor_starter. Each hook is described, including when it is invoked, what task it is supposed to accomplish, what data is passed to the hook, what output is expected, and, when relevant, the exit status expected.
The hook defined by the configuration variable <Keyword>_HOOK_FETCH_WORK is invoked whenever the condor_startd wants to see if there is any work to fetch. There is a related configuration variable called FetchWorkDelay which determines how long the condor_startd will wait between attempts to fetch work, which is described in detail in within section 4.4.1 on page 1371. <Keyword>_HOOK_FETCH_WORK is the most important hook in the whole system, and is the only hook that must be defined for any of the other condor_startd hooks to operate.
The job ClassAd returned by the hook needs to contain enough information for the condor_starter to eventually spawn the work. The required and optional attributes in this ClassAd are identical to the ones described for Computing on Demand (COD) jobs in section 4.3.3 on COD Application Attributes, page 1336.
The hook defined by the configuration variable <Keyword>_HOOK_REPLY_FETCH is invoked whenever <Keyword>_HOOK_FETCH_WORK returns data and the condor_startd decides if it is going to accept the fetched job or not.
The condor_startd will not wait for this hook to return before taking other actions, and it ignores all output. The hook is simply advisory, and it has no impact on the behavior of the condor_startd.
The hook defined by the configuration variable <Keyword>_HOOK_EVICT_CLAIM is invoked whenever the condor_startd needs to evict a claim representing fetched work.
The condor_startd will not wait for this hook to return before taking other actions, and ignores all output. The hook is simply advisory, and has no impact on the behavior of the condor_startd.
The hook defined by the configuration variable <Keyword>_HOOK_PREPARE_JOB is invoked by the condor_starter before a job is going to be run. This hook provides a chance to execute commands to set up the job environment, for example, to transfer input files.
The condor_starter waits until this hook returns before attempting to execute the job. If the hook returns a non-zero exit status, the condor_starter will assume an error was reached while attempting to set up the job environment and abort the job.
The hook defined by the configuration variable <Keyword>_HOOK_UPDATE_JOB_INFO is invoked periodically during the life of the job to update information about the status of the job. When the job is first spawned, the condor_starter will invoke this hook after STARTER_INITIAL_UPDATE_INTERVAL seconds (defaults to 8). Thereafter, the condor_starter will invoke the hook every STARTER_UPDATE_INTERVAL seconds (defaults to 300, which is 5 minutes).
The condor_starter will not wait for this hook to return before taking other actions, and ignores all output. The hook is simply advisory, and has no impact on the behavior of the condor_starter.
The additional attributes included inside the job ClassAd are:
The hook defined by the configuration variable <Keyword>_HOOK_JOB_EXIT is invoked by the condor_starter whenever a job exits, either on its own or when being evicted from an execution slot.
The condor_starter will wait for this hook to return before taking any other actions. In the case of jobs that are being managed by a condor_shadow, this hook is invoked before the condor_starter does its own optional file transfer back to the submission machine, writes to the local job event log file, or notifies the condor_shadow that the job has exited.
The job ClassAd passed to this hook contains all of the extra attributes described above for <Keyword>_HOOK_UPDATE_JOB_INFO , and the following additional attributes that are only present once a job exits:
Hooks are defined in the HTCondor configuration files by prefixing the name of the hook with a keyword. This way, a given machine can have multiple sets of hooks, each set identified by a specific keyword.
Each slot on the machine can define a separate keyword for the set of hooks that should be used with SLOT<N>_JOB_HOOK_KEYWORD . For example, on slot 1, the variable name will be called SLOT1_JOB_HOOK_KEYWORD. If the slot-specific keyword is not defined, the condor_startd will use a global keyword as defined by STARTD_JOB_HOOK_KEYWORD .
Once a job is fetched via <Keyword>_HOOK_FETCH_WORK , the condor_startd will insert the keyword used to fetch that job into the job ClassAd as HookKeyword. This way, the same keyword will be used to select the hooks invoked by the condor_starter during the actual execution of the job. However, the STARTER_JOB_HOOK_KEYWORD can be defined to force the condor_starter to always use a given keyword for its own hooks, instead of looking the job ClassAd for a HookKeyword attribute.
For example, the following configuration defines two sets of hooks, and on a machine with 4 slots, 3 of the slots use the global keyword for running work from a database-driven system, and one of the slots uses a custom keyword to handle work fetched from a web service.
The keywords "DATABASE" and "WEB" are completely arbitrary, so each site is encouraged to use different (more specific) names as appropriate for their own needs.
There are two events that trigger the condor_startd to attempt to fetch new work:
Even if a given compute slot is already busy running other work, it is possible that if it fetched new work, the condor_startd would prefer this newly fetched work (via the Rank expression) over the work it is currently running. However, the condor_startd frequently evaluates its own state, especially when a slot is claimed. Therefore, administrators can define a configuration variable which controls how long the condor_startd will wait between attempts to fetch new work. This variable is called FetchWorkDelay .
The FetchWorkDelay expression must evaluate to an integer, which defines the number of seconds since the last fetch attempt completed before the condor_startd will attempt to fetch more work. However, as a ClassAd expression (evaluated in the context of the ClassAd of the slot considering if it should fetch more work, and the ClassAd of the currently running job, if any), the length of the delay can be based on the current state the slot and even the currently running job.
For example, a common configuration would be to always wait 5 minutes (300 seconds) between attempts to fetch work, unless the slot is Claimed/Idle, in which case the condor_startd should fetch immediately:
If the condor_startd wants to fetch work, but the time since the last attempted fetch is shorter than the current value of the delay expression, the condor_startd will set a timer to fetch as soon as the delay expires.
If this expression is not defined, the condor_startd will default to a five minute (300 second) delay between all attempts to fetch work.
The availability of multiple versions of an application leads to the need to specify one of the versions. As an example, consider that the java universe utilizes a single, fixed JVM. There may be multiple JVMs available, and the HTCondor job may need to make the choice of JVM version. The use of a job hook solves this problem. The job does not use the java universe, and instead uses the vanilla universe in combination with a prepare job hook to overwrite the Cmd attribute of the job ClassAd. This attribute is the name of the executable the condor_starter daemon will invoke, thereby selecting the specific JVM installation.
In the configuration of the execute machine:
With this configuration, a job that sets the HookKeyword attribute with
in the submit description file causes the condor_starter will run the hook specified by JAVA5_HOOK_PREPARE_JOB before running this job. Note that the double quote marks are required to correctly define the attribute. Any output from this hook is an update to the job ClassAd. Therefore, the hook that changes the executable may be
If some machines in your pool have this hook and others do not, this fact should be advertised. Add to the configuration of every execute machine that has the hook:
The submit description file for this example job may be
Note that the requirements command ensures that this job matches with a machine that has JAVA5_HOOK_PREPARE_JOB defined.
Job Router Hooks allow for an alternate transformation and/or monitoring than the condor_job_router daemon implements. Routing is still managed by the condor_job_router daemon, but if the Job Router Hooks are specified, then these hooks will be used to transform and monitor the job instead.
Job Router Hooks are similar in concept to Fetch Work Hooks, but they are limited in their scope. A hook is an external program or script invoked by the condor_job_router daemon at various points during the life cycle of a routed job.
The following sections describe how and when these hooks are used, what hooks are invoked at various stages of the job’s life, and how to configure HTCondor to use these Hooks.
The Job Router Hooks allow for replacement of the transformation engine used by HTCondor for routing a job. Since the external transformation engine is not controlled by HTCondor, additional hooks provide a means to update the job’s status in HTCondor, and to clean up upon exit or failure cases. This allows one job to be transformed to just about any other type of job that HTCondor supports, as well as to use execution nodes not normally available to HTCondor.
It is important to note that if the Job Router Hooks are utilized, then HTCondor will not ignore or work around a failure in any hook execution. If a hook is configured, then HTCondor assumes its invocation is required and will not continue by falling back to a part of its internal engine. For example, if there is a problem transforming the job using the hooks, HTCondor will not fall back on its transformation accomplished without the hook to process the job.
There are 2 ways in which the Job Router Hooks may be enabled. A job’s submit description file may cause the hooks to be invoked with
Adding this attribute to the job’s ClassAd causes the condor_job_router daemon on the submit machine to invoke hooks prefixed with the defined keyword. HOOKNAME is a string chosen as an example; any string may be used.
The job’s ClassAd attribute definition of HookKeyword takes precedence, but if not present, hooks may be enabled by defining on the submit machine the configuration variable
Like the example attribute above, HOOKNAME represents a chosen name for the hook, replaced as desired or appropriate.
There are 4 hooks that the Job Router can be configured to use. Each hook will be described below along with data passed to the hook and expected output. All hooks must exit successfully.
The hook defined by the configuration variable <Keyword>_HOOK_TRANSLATE_JOB is invoked when the Job Router has determined that a job meets the definition for a route. This hook is responsible for doing the transformation of the job and configuring any resources that are external to HTCondor if applicable.
The hook defined by the configuration variable <Keyword>_HOOK_UPDATE_JOB_INFO is invoked to provide status on the specified routed job when the Job Router polls the status of routed jobs at intervals set by JOB_ROUTER_POLLING_PERIOD .
The hook defined by the configuration variable <Keyword>_HOOK_JOB_FINALIZE is invoked when the Job Router has found that the job has completed. Any output from the hook is treated as an update to the source job.
The hook defined by the configuration variable <Keyword>_HOOK_JOB_CLEANUP is invoked when the Job Router finishes managing the job. This hook will be invoked regardless of whether the job completes successfully or not, and must exit successfully.
The Daemon ClassAd Hook mechanism is used to run executables (called jobs) directly from the condor_startd and condor_schedd daemons. The output from these jobs is incorporated into the machine ClassAd generated by the respective daemon. This mechanism and associated jobs have been identified by various names, including the Startd Cron, dynamic attributes, and a distribution of executables collectively known as Hawkeye.
Pool management tasks can be enhanced by using a daemon’s ability to periodically run executables. The executables are expected to generate ClassAd attributes as their output; these ClassAds are then incorporated into the machine ClassAd. Policy expressions can then reference dynamic attributes (created by the ClassAd hook jobs) in the machine ClassAd.
The output of the job is incorporated into one or more ClassAds when the job exits. When the job outputs the special line:
the output of the job is merged into all proper ClassAds, and an update goes to the condor_collector daemon.
As of version 8.3.0, it is possible for a Startd Cron job (but not a Schedd Cron job) to define multiple ClassAds, using the mechanism defined below:
In other words, the syntax is:
- [name] [update: bool]
For example, if the Startd Cron job returns:
it will set Value=10 for all slots except slot1 and slot2. On those slots it will set Value=1 and Value=2 respectively. It will also send updates to the collector immediately.
Configuration variables related to Daemon ClassAd Hooks are defined in section 3.5.31.
Here is a complete configuration example. It defines all three of the available types of jobs: ones that use the condor_startd, benchmark jobs, and ones that use the condor_schedd.