Work the Processes page
- Synopsis
- Full Details
In briefProcesses (Operations › Processes) is the live view of the run in the selected environment. Job View lists every job, Workflow View lists the workflow instances, and from either you act on one row through its action menu or on two or more through Bulk actions. Selecting a job opens its detail panel: Output, Summary, Configuration, Status History and Status Update.
- A job's action menu lists every action and disables the ones its status refuses; Bulk actions offers only what every selected row accepts.
- Job Actions in a workflow instance's action menu changes every job in that instance whose status allows the action, and reports the rest.
- Show jobs with errors covers Failed, Marked Failed, Initialization Error, Missed Start Time, Under Review and Lost in Failover.
- In Workflow View, a red "
Nfailed" beside a status, or Completed with errors, marks a workflow with failed jobs. There is no filter for it. - Schedule Workflow builds a date range of up to 31 days; Job Execution History looks back up to 366 days.
- View Workflow History, on an open Workflow View row, lists that workflow's finished runs across days.
- An overrun of Max Run Time is a badge, not a status change, and it stays after the job ends.
- Changing the page, sort, search or a filter clears your selection.
- Hold, Release or Start on a completed workflow instance reopens it, but nothing runs until you restart a job on it.
- A run reopened with Start can't be deleted; use Close to put it back to Completed.
- The platform does not stop a job that overruns its Max Run Time.
Processes is your daily-run screen: the live view of the workflows and jobs running in the selected environment. This page is where you watch the run and act on it.
During the run you need to see what's happening and act fast (push a waiting job through, stop one that's misbehaving, or clear a batch of stuck jobs at once) without hunting through separate screens or selecting one row at a time.
- Views
- Finding Jobs
- Reading a Job
- Acting on Jobs
- Reopening a Workflow
- Adding a Job
Views
Pick your view
A toggle switches between two views (your choice sticks if you reload):
- Job View (the default): every job in the run, one row each. All columns sort.
- Workflow View: the workflows themselves. Filter it by status group (Building, Waiting, In
Process, On Hold, Completed, Cancelled) rather than individual statuses. A red "
Nfailed" beside a workflow's status says that many of its jobs failed, and a finished workflow with failed jobs reads Completed with errors rather than Completed. See Workflows with failed jobs.
From here you can also Schedule Workflow to add a workflow to the run, and Refresh the view. Refresh also reloads an open job's detail and the jobs in an expanded workflow row. Schedule Workflow takes a date range of up to 31 days, not just one day, and builds each date in turn — so a date the workflow isn't deployed for is reported against that date while the rest of the range still builds. One tick of Overwrite Existing applies to every date you picked; the confirmation states the date count beside the instance count, and carries the progress while the range runs.
Opening a row lifts it to just under the column headers and scrolls it into view, and its panel gets the full height of the list — so a workflow diagram has room instead of being squeezed into whatever the other rows left over. The rest of the rows carry on below the panel in their usual order, and collapsing the row puts it back where it was. Your sort and filters don't change.
The list refreshes itself every 30 seconds. While a row is open it pauses, so new jobs starting can't push your row off the page, and a note above the table says so. The open row's diagram and job detail keep updating on their own, and your actions on a row still show at once. Close the row and the list refreshes straight away, then carries on every 30 seconds.
Following a job between the views
Switching views with the tab strip keeps you on the same days: the Date(s) filter, or a Choose Hours preset, goes with you, while the name, Status(s) and Tag filters stay with each view. Opening a row and closing it again returns the list to where you had scrolled.
Both views hand you off to the other with the destination already narrowed:
- On the workflow diagram, a job node's menu offers View in Job View. It switches views, fills in the Job, Workflow and Date(s) filters from that job, and opens that job's own row — so you land on the job you clicked, not on every job that shares its name. Earlier builds seeded only the name, which matches anywhere in a name that repeats across workflows.
- From a job's detail, Workflow → goes the other way: the Workflow View filters to that job's workflow instance and opens its row.
Editing the seeded Job filter — including clearing it to widen the list — hands the view back to you and abandons the rest of the hand-off, so the filter bar always matches what you're looking at.
Finding Jobs
Find what you're looking for
- Type in the Job column header filter to find a job in the run by name.
- Use the filter bar across the top (Choose Hours, Date(s), Status(s) and Saved Filters, plus Tag in Job View) and the filter box in the Workflow, Job and Agent column headers to narrow the list.
- Filter by status group rather than by raw status. In Job View the Status(s) list offers On Hold, Waiting, Running, Processing Events, Finished OK, Fixed, Skipped, Failed, Cancelled, Under Review and Lost in Failover, in one list — the names the status badges use. Earlier builds called three of them Held, Pending Term and Succeeded; a saved filter that used one still works.
Fixed and Skipped are now choices of their own. They used to be reachable only inside Finished OK (then called Succeeded), so there was no way to list just the jobs that were skipped. Finished OK still includes both, so it means what it always did.
Narrowing a name filter
The name filters — Workflow, Job and Agent, each in its own column header — all match
anywhere in the name by default, which is usually what you want and occasionally far too much.
Add a * to say where the match has to sit:
| You type | You get |
|---|---|
payroll | Any name containing payroll |
payroll* | Names starting with payroll |
*payroll | Names ending with payroll |
pay*roll | Starts with pay and ends with roll |
Case never matters. ?, _ and % are all taken literally, so a name that really contains one is
findable — earlier builds let _ and % through as wildcards, which is why a filter of JOB_01
used to bring back JOB-01 as well.
The same applies to the Job History report's workflow, job and agent name filters. See The Processes page.
To see only what needs attention, turn on Show jobs with errors. It covers every failure status — Failed, Marked Failed, Initialization Error, Missed Start Time and Under Review, including the brief "pending" form of Failed, Marked Failed and Under Review — and Lost in Failover, which needs you to record the outcome just as an Under Review job does. Earlier builds matched only Failed and Under Review, so a job that failed to initialize or missed its start window stayed hidden while the switch was on; if you had stopped trusting it, it is reliable now.
Changing the page, sort, search, or a filter clears your current selection, so a bulk action never acts on rows you can no longer see.
Telling two runs of the same workflow apart
Some workflows build more than once for the same date — once per named schedule instance, such as
one run per branch. Those rows show a composed name, PAYROLL_BRANCH1, so you can tell them apart.
A workflow that builds only once shows its plain name.
The Workflow filter in the Workflow View's column header matches the composed name, so typing
BRANCH brings back PAYROLL_BRANCH1 and PAYROLL_BRANCH2. There is no longer a separate
Instance Name filter — it used to sit beside Workflow Name and hid every workflow that had no
instance name at all, which made an empty grid hard to explain. A saved filter or bookmark that
carried one still loads; the instance name in it is simply ignored.
One instance can also be built more than once for the same date — when the build carries
different instance property values, or on any build of a workflow that allows multiple instances but
declares none. Those rows carry a four-digit run suffix — PAYROLL_BRANCH1$0002 is the second run of
BRANCH1. The first run has no suffix, and the suffix follows through into the Close and Delete
confirmations and the result message after a bulk action, so an action on one run is never reported
against another.
An underscore means a schedule instance — a whole separate run of the workflow
(PAYROLL_BRANCH1). A $nnnn suffix means a repeat run of that same instance for the date. A
dot means a job that fanned out into several instances inside one run (EXTRACT.US). You may
see all three in the same run.
Reading a Job
Spot a container job with nothing to run
A Workflow Container job runs a whole other workflow as one step. If that embedded workflow couldn't be built — usually because it isn't deployed for the date — the container has nothing to run, and it will fail when it starts. Until then it sits in an ordinary waiting status, which is why it's marked for you:
- A warning icon sits beside the job name in Job View, and beside the status badge on the workflow diagram. Hover it for the reason the build recorded.
- Scheduling a workflow that leaves a container empty raises a warning banner above the list naming each one — the schedule instance that owns it, the container job, the workflow it referenced, and the reason. It has no timer, so it waits for you to dismiss it.
Fix what the reason names — most often, have the embedded workflow deployed for that date — then rebuild the parent with overwrite. If you'd rather let the run go on without it, Skip the container job (or Cancel it, or Mark Finished OK if the dependents should proceed as though it succeeded). Any of those three clears the warning as well.
Spotting a job that overran its limit
A job that has run longer than its Max Run Time is marked rather than failed — its status does not change, so the badge is the only sign. It is shown in three places, and you do not have to open the job to see it:
- Beside the job's status in the Status column of the jobs list.
- As a timer icon on the job's card on the workflow diagram — hover it for Exceeded Max Run Time.
- On the detail panel's Summary tab, with a Max Run Time Exceeded row giving the time it was detected.
The mark stays after the job finishes, including a job that overran and then finished OK — which is exactly the case its status cannot tell you about. It is hidden while you scrub the timeline, because the statuses shown there are reconstructed for that moment and the overrun is a live fact with no history behind it.
The platform does not stop an overrunning job. See Exceeding Max Run Time for what to do if you want it stopped.
See what a job is waiting on from outside this run
A job can be waiting on a predecessor that isn't part of the workflow instance you are looking at — a job in another workflow, or a job in this workflow on a different date. The workflow diagram used to draw nothing at all for those, so the waiting job appeared to have no reason to be waiting.
That predecessor is now a dimmed stand-in card with an arrow into the job that waits on it. Read it like this:
- the top line is the workflow the predecessor belongs to, and the bold line its job name;
- a trailing
*on the name means the dependency matches a name prefix, so more than one job could satisfy it; - a third line names the schedule instance, when the dependency names one;
-1dor+2dis the day relative to the waiting job's own date — blank means the same day;Any daymeans the relation isn't tied to a single date.
The card carries no status, because nothing here checks that other run. The arrow is what tells you whether the wait is still on: a satisfied dependency draws solid, an unsatisfied one dashed.
Nothing on this canvas will resolve it. Go and look at the run the card names — switch the Date filter to the day it points at, or search for that workflow — and act there.
Read a job's output
The Output tab on a job's detail panel shows what the job printed, with the job's Date, Exit Code and Workflow above it so a log you copy elsewhere still says which run it came from.
What the tab shows depends on whether the run has an LSAM job number, which a legacy (LSAM) run gets shortly after its agent accepts the job, while it is still running. A run that has one gives you a list of output files the agent holds; open any one to read it in place. A run without one — a job on a Continuum agent, or a legacy run not yet dispatched or just restarted — gives you the one log returned with the job, with Refresh output to read it again. A job that runs no process of its own — a container job, for instance — has no output and says so.
Four things are worth knowing when the log is the thing you actually need:
- Open shows the file you are reading in a new browser tab, at full width. The in-panel view materialises only the first 500 KB of a long file and tells you when it has — Use Open to read all of it. Open and Download always work from the whole file.
- Download never truncates. All takes every output file; the list beside it takes one named file.
- The files are listed in the order the agent sent them. A z/OS job's SYSOUT is shown as the mainframe lays it out: each row keeps its spacing, and a wide line scrolls sideways rather than wrapping. If its output cannot be listed, the tab says why — see Its output. An MCP job's output is listed while it runs as well as after, and an OS 2200 job's once its agent reports where it is — see MCP and OS 2200.
- Refresh from agent re-asks the agent for the file list. While the job is running, every look asks the agent, so you see the output as it stands; once it has ended, the list is fetched once and then kept. A rerun needs no refresh — the output you see belongs to the run you are looking at, not the one before it. There is no automatic refresh otherwise; it is yours to ask for.
Retrieving output from a UNIX or IBM i agent depends on the relay being on a current build. On an older relay, a UNIX job's Output tab came back holding the agent's own global log files instead of the job's output, reporting success the whole way — so it looked right and was wrong. The platform now discards files that don't belong to the job before storing a UNIX listing, so on an older relay new output comes back empty instead. If UNIX or IBM i output is empty or clearly belongs to something else, ask your Administrator to check the relay build before treating it as a job fault. Once the relay is current, use Refresh from agent on any job whose wrong list was already stored.
Acting on Jobs
Act on one job
Open a job's action menu to act on it. The menu lists every action and disables the ones the job's status does not accept, so you can't pick one that would fail; the Status Update tab on its detail panel lists only the ones it accepts. For what each action does, including when to use Kill vs Cancel, see Job statuses and actions.
Select a job to open its detail panel (read-only). It has five tabs — Output, Summary, Configuration, Status History and Status Update — and a link to the job's Workflow.
Status History is where the job's status changes live. It used to be a dialog opened from an icon in the panel header; that icon is gone, and the tab shows the same thing — each change with its timestamp, the old and new status, who made it and any detail. It loads only when you open the tab.
To compare tonight against other nights, use the Job Execution History button on that same tab. It lists this job's runs over a date range — start, end, runtime, agent, termination status and exit code — so you can tell a job that is slow tonight from one that is always slow. A chart above the list plots each run's run time across the dates you chose (up to 366 days), so a job that is getting slower shows as a rising line. Opening a run there gives you its recorded detail and a Notes tab, which is the place to leave what you found for whoever looks next. Notes belong to the job instance, so every run of it on that date (a restart, a retry) shows the same notes; the same job in another workflow instance has notes of its own. See Earlier runs of the same job.
To compare a whole workflow run against other nights, open its row in Workflow View and select View Workflow History on the workflow details panel beside the diagram. It lists the workflow's finished runs, Completed or Cancelled, over the last 30 days by default — date, start, end, run time, status, how many jobs it had and how many failed — newest first. For a named schedule instance it lists only that instance's runs. See Earlier runs of a workflow.
Restart is in Status Update, with the other actions, rather than in a footer button.
Act on many jobs at once (bulk actions)
To act on several jobs at once, complete the following steps:
- Select the rows you want, or select them all.
- Once two or more are selected, a Bulk actions button appears.
- Open it. The menu shows only the actions valid for every selected row, plus Delete when every selected row can be deleted.
- Choose an action; it runs across the selection and a message confirms the result.
This works in both Job View and Workflow View.
Act on every job in one workflow instance
To hold, release, restart, cancel, skip or mark every job in one workflow instance, complete the following steps:
- In Workflow View, open the instance's action menu and choose Job Actions.
- Choose the action. A Change Job Status dialog says how many jobs will change, and how many won't because their status doesn't allow it.
- Read the count, and confirm. The dialog stays open until every job has been tried.
- Read the result. It says how many jobs changed, and why any did not: their status, a status that changed in the meantime, another action in progress on them, or a missing permission.
Only the instance's own jobs are changed, not the jobs inside a container's nested sub-workflow. Force Start, Kill and Delete aren't offered here. On a Completed instance, Hold, Release and Restart reopen it, and the confirmation says so; on a completed nested sub-schedule they're not available. See Job actions on a whole workflow instance.
Delete a job or workflow instance
Delete removes a job or a whole workflow instance from the run (a soft delete). Deleting a job also frees anything waiting on it, so downstream jobs stop hanging on a step that's no longer there.
Delete isn't in a job's action menu or its Status Update tab. Delete a job with Delete Job on its card in the workflow diagram, or with Bulk actions; delete a workflow instance from its row menu in Workflow View, or with Bulk actions.
- You can't delete a job while it's running or in a pending-termination status. Kill it first.
- You can't delete a workflow instance while it's actively running, or while one of its jobs is. Hold or let it finish first. Deletion is allowed when it's waiting, held, completed, or cancelled.
- You can't delete a nested sub-schedule on its own; it follows its parent workflow.
Reopening a Workflow
Reopen a workflow that already finished
A completed workflow instance isn't the end of the road. From Workflow View, Hold, Release and Start are all available on a Completed row, and each one reopens that finished run so you can get more work out of it:
| Choose | And the run's row then reads |
|---|---|
| Hold | On Hold |
| Release | Wait To Start |
| Start | Started By User |
A run finished, and then you find out a job in it has to run again — a step that failed and was marked fixed too early, or output that has to be regenerated. You need that same instance back, not a whole new build for the date.
To reopen a completed run and re-run a job in it, complete the following steps:
- In Workflow View, open the completed instance's action menu and choose Hold, Release or Start. Release is the usual choice — it puts the run back in the queue.
- Confirm in the Reopen Completed Workflow dialog. It names the instance and tells you where the action will leave it. Choose Leave Completed if you've changed your mind.
- Restart the job you reopened the run for. This is the step that matters — see the note below.
Reopening clears the run's recorded End Time and moves it out of Completed, but it does not restart anything. The instance sits in its new status and stays there — it won't run on, and it won't quietly re-complete itself either — until you restart a job on it. Once you do, it picks up and finishes as a genuine second run, and its completion notifications fire again rather than being treated as a repeat of the first.
If you reopened with Hold, remember it's now a held run: release it as well.
Two things to know:
- You can't reopen a sub-schedule. On a completed nested sub-schedule (a run started by a container job), those three actions aren't available — a sub-schedule follows its parent. Act on the parent workflow instead.
- A run reopened with Start can't be deleted. Started By User isn't a deletable status, so if you reopen with Start and then change your mind, use Close to put the run back to Completed. Reopening with Hold or Release leaves you a run you can still delete.
- Restarting a job on the completed run reopens it too. If you restart or force-start a job on a Completed run without reopening it first, the run goes back to In Process and the job runs — even for a run that completed before the platform last restarted.
To reopen several runs at once, select them and use Bulk actions. If any selected row is Completed, the action confirms first and tells you how many completed runs it will reopen — worth reading on a mixed selection.
Once a run has started — including a run you reopened — the automatic daily build leaves it alone instead of rebuilding over it. That's deliberate: it won't overwrite work that has already happened. Rebuild the date yourself if that's what you want.
Adding a Job
Add a job to a running workflow
Sometimes a workflow is already running (or has finished for the day) and you need one more of its jobs to run: a step that was skipped, or one you want to run again. From Workflow View, open a workflow instance's action menu and choose Add Job.
A job that belongs to the workflow didn't run this time (maybe its frequency didn't qualify for the date, or it was left out), and you need it to run now, in the same instance, without rebuilding or rescheduling the whole workflow.
To add a job to a running workflow, complete the following steps in the Add Job to Workflow Instance dialog:
- Pick the job. You'll see every job defined on the workflow. Each is tagged so you know what will
happen:
- (no tag): the job isn't in this run yet and will be added.
- Will Replace: the job already ran and finished; adding it again replaces that finished copy (you'll be asked to confirm first).
- Will Add Copy: an instance of a multi-instance job is still running or waiting, and adding it puts a numbered copy beside it. You can select it — see the note below.
- Job is Active (greyed out): the job is already running or waiting, so you can't add another. This is for a job that is not multi-instance; a multi-instance one gets a copy instead.
- No Frequency (greyed out): the job has no frequency defined, so it can't be added.
- Disabled: the job's Disabled switch is on, so no build ever includes it. You can still add it — that is the way to run a parked job once — but the badge is there so it is a choice.
- Choose a frequency if the job has more than one (if it has just one, it's picked for you).
- Choose which instances — only for a job that runs as multiple instances. Tick All Instances to add every one of them at once, or leave it clear and pick a single Instance.
- Adjust properties if you need to. The grid is pre-filled with the job's defaults; leave it alone to use them as-is.
- Add on Hold if you want the job held instead of released to run right away.
- Add a reason (optional). It's saved to the job's history for the audit trail.
Choose Add Job. A released job starts as soon as it's eligible; a held job waits until you release it. If you add a job to a workflow that already completed, the workflow reopens so the new job can run. The added job shows up in the run and on the workflow's timeline, marked as added.
Each instance is handled separately, which is usually what you want: one of them already running doesn't stop you adding the others, and if one can't be added the rest still are. The result tells you what happened to each.
Its tag is read per instance, too — from each instance's own row, not from the plain job name, which is never in the run. An instance still running reads Will Add Copy; one that has finished makes it a Will Replace. Earlier builds looked under the plain name, found nothing, and showed every multi-instance job as freely addable — so Will Replace never appeared and a finished instance could be replaced without a confirmation.
This used to be refused — already active in this instance — which meant you could not ask for a
second run of an instance while the first was still going, even though that is exactly what a
multi-instance job is for. Now you get a numbered copy beside the live one: EXTRACT.US$0001
next to EXTRACT.US.
Two things to expect:
- The copy behaves as its instance. Anything that waits on
EXTRACT.USwaits on its copies too, and the copy waits on whateverEXTRACT.USwaits on. You are not creating an orphan. - A finished copy is reused before a new number is taken, lowest first, so the numbers do not
climb forever. Because that replaces the finished copy, the job reads Will Replace rather than
Will Add Copy, the warning names the copy, such as
EXTRACT.US$0001, and you are asked to confirm. Earlier builds replaced it without a word. Past$9999that one instance is refused and the rest of an All Instances add still goes through.
A job that is not multi-instance still reads Job is Active and cannot be added twice. See A live instance gets a copy, not a conflict.
A $JOB:ADD or $JOB:ADDHLD event names a job but has no field for an instance, so a multi-instance
job used to be refused outright. Now, with no properties on the event it adds every instance,
and with properties it adds one, named from the first
property's value — so Property=1 is its own instance beside Property=0, which is the point of a
multi-instance job. If that value happens to name one of the job's own instances, that instance
is added instead, with the event's properties merged over its own.
Either way the event log records what happened to each instance, and a partly failed add is worth reading before you raise it again — see Use the event log.
A job assigned to a legacy agent group in run-on-all mode used to be refused. You can add one
now, and it behaves as the build does: you get one job per member of the group, each named
<job>__<agent> and pinned to that agent, each reporting its own result.
The group is read as it stands now, not as it was when the workflow was built, so an agent added to the group since then is included. Two things to expect:
- The dialog reads the job's tag from its per-agent jobs, because there is no single job under the
plain name. Every agent still running it reads Job is Active; any agent that has finished makes
it a Will Replace. On a job that is also multi-instance, copies are numbered per member —
EXTRACT.US$0001__AG1— and one member's copies don't affect another's numbering. - The add is refused, naming the group, if the group has no members, if two members share a name, or if the group can't be read.
A Universal Agent pool set to run on all its agents is added as a single job instead — again, the same as the build.
You can't add a job to a cancelled workflow instance.
Workflow-container (sub-schedule) jobs can be added now, with one restriction: if you're replacing a container job that already ran, its child workflow has to be finished first. Replacing one while the child is still going would leave that child with nothing owning it, so it's refused rather than allowed.
Related topics
- The Processes page (runtime) — full configuration reference and troubleshooting
- Monitor the daily run (Operator)
- Job statuses and actions (Operator)
- Respond to a failed job (Operator)