The Processes page (runtime)
- Synopsis
- Full Details
In briefProcesses is the daily-run screen: the live view of workflow instances and their jobs for the selected environment. It shows the run as Job View or Workflow View, and is where you read each job's detail and output. What you can do to the run (schedule, act, add and edit jobs) is on Acting on runs in Processes.
WhereOperations › Processes. It shows the environment selected in the environment picker above it. A tab strip switches between Job View, the default, and Workflow View. A job's earlier runs are on its detail panel: Status History › Job Execution History. A workflow's earlier runs are behind View Workflow History on the workflow details panel of an open Workflow View row.
When
- You are watching the run and need to see what is waiting, running, held or failed. In Workflow
View, a red "
Nfailed" beside a workflow's status, or Completed with errors, says it has failed jobs. - You want to read a job's output, or see across up to 366 days whether it is getting slower.
- You want to see how a workflow's earlier runs went: when each started and ended, and how many of its jobs failed.
Use something else when you need to act on the run: see Acting on runs in Processes. When you need to know what a status or action means, see Job and workflow statuses and Operator response guide. For what a threshold or resource stands at right now, use the Toolkit.
How
- Open Operations › Processes, and check the environment above the page.
- Choose Job View or Workflow View.
- Narrow the list with the filter bar (Date(s), Status(s), Saved Filters, and in
Job View Tag) or a
column header filter. Name filters accept
*. - Select a job to open its detail panel: Output, Summary, Configuration, Status History and Status Update.
For the full steps, see Work the Processes page.
- The view you choose is kept in the page address, so a reload returns to it.
- The list refreshes itself every 30 seconds, except while a row is open. Closing a row returns the list to where you had scrolled.
- Switching between Job View and Workflow View carries the date filter across; the other filters stay with their view.
- Refresh also reloads an open job's detail, and a running job's Duration counts up every second.
- Notes in Job History Detail belong to the job instance, so every run of it shows them.
- Workflow History lists only finished runs, Completed or Cancelled, and at most the 100 most recent in the range.
- Each action needs its permission in the workspace of the workflow the instance was built from. Reading needs only the view permission for that list (Workflow Instances View for Workflow View, Job Instances View for Job View), in any workspace of the environment.
- A deleted note cannot be recovered.
- If UNIX or IBM i output comes back wrong or empty, check the relay build first.
See Job and workflow statuses and Operator response guide. A status is spelled on Processes as its badge shows it — Wait Start Time, Job Running — which differs from the name the status reference uses; see How each screen spells a status.
Task walkthrough: Work the Processes page. This page is the full configuration and troubleshooting reference.
Processes is the operator's daily-run screen: the live view of workflow instances and their jobs for the selected environment. It is environment-scoped (the environment picker sits above it) and breadcrumbs to a single Processes crumb.
Two views
A view toggle switches the body between two views; the choice is URL-synced, so a reload returns to the same view. Job View is the default.
The toggle is a tab strip, and it takes the keyboard: Tab reaches the selected view, then the arrow keys, Home and End move between the two and select as they go. The same is true of every other tab strip in the product, the job detail panel's tabs included.
| View | Shows | Notes |
|---|---|---|
| Job View | Every job instance in the run | Default. All column headers sort. |
| Workflow View | Workflow instances (the containers) | Filters by coarse status groups, not raw statuses (see below). "View in Workflow view" links from a job. |
The page also hosts the Schedule Workflow action (schedule a workflow into the run) and a manual Refresh. The dialog's Workflow list is sorted alphabetically by name, so you can scan it for the one you want.
Refresh reloads the list, and with it an open job detail panel and the job lists inside expanded workflow rows, so their statuses and durations are current too. Your filters stay as they were. Earlier builds reloaded only the list, so an open job's detail kept its old status.
An expanded row moves to the top while it is open. Opening a row lifts it directly under the column headers, scrolls the list up so it is in view, and gives its panel the full height of the visible table — the rest of the rows continue below the panel in their usual order, and collapsing the row returns it to its place. Earlier builds sized the panel to whatever space the other rows left over, so on a full page a workflow diagram was squeezed to its minimum and changed height as the row count changed.
Only the display order moves. Your sort and your filters are untouched, so the row order you see below the panel is the order you chose.
The row menu stays on screen when the table scrolls sideways. In a narrow window both views scroll their columns sideways. The expand and selection columns and the first column, in the order you have arranged them (Date by default), stay pinned on the left, and the ⋮ row-action column stays pinned on the right, so a row's actions are always in reach. Earlier builds scrolled the ⋮ menu off screen with the rest of the columns.
Closing the row puts the list back where you were. The list remembers how far down you had scrolled when you opened the first row, and returns there when you close the last one. Changing the page, sort, filters, search or page size in between forgets that position, because it belongs to a different list. Earlier builds left the list scrolled to the top after you closed a row.
A job you restart while its row is open stays open. A restart clears the job's start time, so under a start-time sort the job can drop off the page until its new run starts. The open row is held on screen in the meantime, showing the job's current detail, priority and tags. Once the job starts again, or is skipped or marked Finished OK or Failed without starting, the list is reloaded: if the job is back on the page, its row stays open where your sort and filters place it; if not, the row closes and leaves the selection. Only a restarted row is held. An open row for a job that has never run, and that leaves the page because your filter no longer matches it, closes at once. Earlier builds kept a restarted row held after it was skipped or marked, and could keep other open rows that left the page.
The list refreshes itself every 30 seconds, and pauses while a row is open. In both views, an open row stops the list's automatic refresh, including the one that runs when you come back to the browser window. A note above the table reads "Auto refresh is paused while a row is open. Close the row to resume." Closing the last open row refreshes the list once and resumes the 30-second refresh. While the list is paused, the open row keeps its own content current: the workflow diagram every 10 seconds, and a running job's detail every 15 seconds. An action you take on a row, such as Hold or Restart, still updates the list. Changing the page, sort or filters closes the open row, as before. Earlier builds kept refreshing, so as new jobs started, an open row could be pushed off the page and close, or vanish from Workflow View.
To act on the run, see Acting on runs in Processes: scheduling a workflow, the row and bulk actions, deleting, adding a job to an instance that is already running, and editing a job on a run under way.
- Reading the Page
- Dependencies
- Filters and Search
- Job Detail
- Output
- Permissions
Reading the Page
Getting around the diagram
Both views render the same run diagram, and its controls are grouped in one panel at the bottom-right — the minimap, with the view controls in a row directly beneath it. Earlier builds scattered them: the zoom controls bottom-left, the minimap bottom-right, and the orientation toggle as a third overlay of its own.
| Control | What it does |
|---|---|
| Minimap | The whole run in miniature, with your current viewport marked. Drag it to pan and scroll it to zoom, so you can cross a large run without touching the canvas. |
| Zoom in / Zoom out | Step the zoom level. |
| Fit view | Zoom and pan so the whole diagram is on screen. |
| Orientation | Switches the layout between top-to-bottom and left-to-right. |
Every button carries a tooltip, and every control is reachable by assistive technology.
The Job view diagram dims every card except the one you are focused on. Those inactive cards are now solid rather than see-through: earlier builds drew them translucent, which let the dependency lines behind them show through the card and made both hard to read. A failed card keeps its error ring while inactive, greyed with the rest of the card rather than staying full-strength red.
Reading a named schedule instance
When a workflow builds as several named schedule instances, both views show
the composed name Workflow_Instance — PAYROLL_BRANCH1 — in the workflow column, so two branches
of one workflow on one date are distinguishable at a glance. A workflow that builds no named instance
shows the bare workflow name.
- The Workflow View has no Instance Name filter. To find one instance, filter on the workflow name and read the composed name in the Workflow column.
- Sorting by the Workflow column orders by the bare name, so an instance's siblings sort as peers.
- Following a job back to its workflow filters the Workflow View to that workflow and date and opens the instance you came from. Sibling instances on that date stay listed.
- The workflow details panel shows an Instance Name row only for a named instance.
- An instance built more than once for the date carries a four-digit run suffix —
PAYROLL_BRANCH1$0002for its second run. The first run has no suffix. The suffix appears everywhere the composed name does, including the Close and Delete confirmations and the result message after a bulk action, so an action on one run is never reported as an action on another.
A schedule instance is named with an underscore (PAYROLL_BRANCH1). A job that fans out into
several instances within a run is named with a dot (EXTRACT.US). Both can appear in the same run.
Workflows with failed jobs
A failed job does not change its workflow's status: the workflow stays In Process while other jobs still run, and becomes Completed once every job has reached a final status, whether or not any of them failed. So the Workflow View's Status column also reads the workflow's jobs:
- A red "
Nfailed" indicator sits beside the workflow's status whenever at least one of its jobs has failed —2 failed, for example. It shows on any workflow status, so a workflow still In Process tells you it already has failures. - A completed workflow with failed jobs reads "Completed with errors", in red, instead of the green Completed. A workflow whose jobs all finished without failing still reads Completed.
A job counts as failed when its status is Failed, Marked Failed, Initialization Error or Under Review, or the pending form of Failed, Marked Failed or Under Review. Missed Start Time is not counted. The count is taken again from the jobs within a few seconds of any job status change, including one the platform makes on its own, so it does not go stale while a workflow runs. The Status(s) filter has no choice for workflows with failed jobs: it still filters by the workflow's own status. To list the failed jobs themselves, use the Job View's Show jobs with errors switch (below).
Earlier runs of a workflow
In the Workflow View, an open row shows the workflow details panel, with its Timeline, Details and Documentation tabs, beside the diagram. View Workflow History on that panel opens Workflow History: the finished runs of this workflow across dates, for the question the run in front of you cannot answer — does this workflow usually finish at this time?
| Part of the dialog | What it shows |
|---|---|
| Workflow Name | Fixed to the workflow you opened it from, in its composed Workflow_Instance form for a named instance, and read-only. |
| Date(s) | A single date or a range. It opens on the 30 days ending on the instance's schedule date. When you pick a range on the calendar, nothing loads until you have chosen both ends. |
| Columns | Date, Instance Name (for a named instance only), Start Time, End Time, Run Time, Status, Jobs and Failed Jobs, newest date first, and paged. |
- Only finished runs are listed: those that ended Completed or Cancelled. A run still in progress, today's included, is not listed until it finishes.
- A named instance lists only its own runs.
PAYROLL_BRANCH1does not listPAYROLL_BRANCH10. A workflow that builds no named instance lists only its runs with no instance name. - Failed Jobs counts the jobs that ended Failed, Marked Failed, Initialization Error
or Missed Start Time. That differs from the red "
Nfailed" indicator above, which counts Under Review and not Missed Start Time. - It shows at most the 100 most recent runs in the range. Past that, the dialog says Only the 100 most recent runs in this date range are shown. Choose a shorter date range to see older runs.
- A range with no finished runs shows No finished runs of this workflow in this date range. If the runs cannot be loaded you get the reason and a Retry. OK closes the dialog.
To report on every workflow's runs at once, or download them, use the Workflow History report.
Status-group filters
Both views filter by status groups rather than raw statuses, each group expanding to its constituent statuses for the query.
- Workflow View offers Building, Waiting, In Process, On Hold, Completed, Cancelled. Unknown is intentionally ungrouped. See Job and workflow statuses.
- Job View offers On Hold, Waiting, Running, Processing Events, Finished OK, Fixed, Skipped, Failed, Cancelled, Under Review and Lost in Failover, in one list. Fixed and Skipped are new as standalone choices — both were previously reachable only folded into Finished OK, which left no way to list a run's skipped jobs. See Job status-group filter.
The choices use the names the status badges show. Earlier builds called three of the Job View groups Succeeded, Held and Pending Term, and the Workflow View's Held. Only the names changed: each group covers the same statuses, and a saved filter or bookmarked URL that used one still applies it.
Showing only jobs with errors
The Job View's Show jobs with errors switch now covers every failure status: Failed, Marked Failed, Initialization Error, Missed Start Time, Under Review and Lost in Failover, plus the pending-term forms of Failed, Marked Failed and Under Review. Lost in Failover is included because, like Under Review, it needs an operator; see Lost in Failover. It previously matched only Failed and Under review, so a job that failed to initialize, one an operator had marked failed, and one that missed its start window were all invisible while the switch was on — and any failed job was invisible for the second or so it spent in pending-term.
Dependencies
Container jobs with no sub-schedule to run
A Workflow Container job whose sub-workflow couldn't be built during the build is marked with a warning icon — beside the job name in the Job View list, and beside the status badge on the workflow diagram in either view. Hovering it gives the reason the build recorded. The job's status is an ordinary waiting one, so without the marker a container that will run nothing looks exactly like one that will.
Scheduling a workflow that leaves any container empty also raises a warning banner above the list, listing each one with its owning schedule instance and the reason. It has no timer and closes only when you dismiss it.
The marker is not shown while the diagram's timeline is replaying an earlier point in the run: the statuses shown there are reconstructed as of that point, while the reason behind the marker is read live, and the two must not be judged against each other.
See Workflow Container job for what the banner says, which statuses carry the marker, and which actions clear it.
Predecessors outside this run
The workflow diagram is drawn from one workflow instance's own jobs, so a predecessor that is not one of them used to have nowhere to appear — the dependency was real, the job was waiting on it, and the canvas showed a job with no incoming arrow at all.
Such a predecessor is now drawn as a dimmed stand-in card with an arrow into the job that waits on it. Two kinds qualify: a job in another workflow, and a job in this workflow on a different schedule date.
The card carries what the dependency stored, which is all a label needs:
| On the card | Meaning |
|---|---|
| Top line | The workflow the predecessor belongs to. |
| Name | The predecessor's job name. A trailing * means the dependency matches on a name prefix, so more than one job could satisfy it. |
| Third line | The schedule instance the dependency names, when it names one. |
-1d, +2d | The Days Offset — which day's run, relative to the waiting job's own schedule date. Blank means the same day. |
Any day | An any-day conflict — the relation is not confined to one date, which is why no offset is shown. |
A stand-in carries no status badge, deliberately: nothing on this canvas resolves it against the instance it names, so there is no state to show. Whether the waiting job is still blocked is carried by the arrow, not by the card.
Several jobs naming the same predecessor share one stand-in, with an arrow each.
Thresholds, resources and expressions on the diagram
The run diagram draws more than job boxes. A job's threshold, resource and expression dependencies each appear as their own card, with an arrow into every job that depends on them — the same cards the Workflows canvas uses, so the two views agree.
| Card | What it stands for | What the arrow is labelled |
|---|---|---|
| Threshold | One named threshold | The comparison that job requires — ≥ 5 |
| Resource | One named resource | The units that job needs — 2 units, 1 unit, or All units for a dependency that uses the whole resource |
| Expression | One expression | Nothing — an expression is satisfied or not |
One card per threshold name, resource name or expression, shared by every job that depends on it, so a resource ten jobs contend for is one card with ten arrows rather than ten copies.
An arrow is solid green when that job's requirement is satisfied and dashed grey when it is not — the same treatment job-to-job arrows already use. A dependency that was met before the platform restarted still reads satisfied after it, including on a job that is already running.
A job sitting in a threshold, resource or expression wait has all its arrows of that kind drawn unsatisfied, even one whose condition is actually met. A job needing two resources shows both as unsatisfied while it waits, because what the diagram knows is that the job is still held by that kind of requirement, not which one of them is holding it. Over-reporting a block is the safer direction: you are told the job is waiting, and reading both resources is the next step.
Current values are not drawn. The card names the threshold or resource; what it stands at right now is on the Toolkit page, not here.
A job blocked on a dependency that will not resolve on its own
A job waiting on a predecessor that will not arrive on its own used to look exactly like a healthy waiting job. It now carries a yellow warning icon in the bottom-left of its card, and the tooltip — "Waiting on a dependency that will not resolve on its own: …" — names each predecessor and why.
Two cases raise it:
- The predecessor is absent from the run, and the dependency is a Requires — which waits
until an operator adds the predecessor. An After dependency on an absent predecessor proceeds,
so it is not marked. The tooltip reads "
<job>is not in this run". - The predecessor is a multi-instance job that has an instance for this job's group, but that
instance is not in the run yet — which waits until it is added, whatever the dependency type. The
tooltip reads "
<job>has no instance for this job's group in this run yet". See A dependency between multi-instance jobs.
The marker follows the job's status, not just the stored reason, so it clears once the job is past its dependency waits — running, finished, waiting on a machine, or waiting to start. Force starting, cancelling or marking the job does not leave a stale warning behind.
Filters and Search
Filters and search
Each view has a filter bar across the top and a filter box in some column headers. Which filter lives where follows what it narrows: the bar holds filters over the whole run, and a filter that narrows one column sits in that column's header.
| View | Filter bar | Column header filters |
|---|---|---|
| Job View | Choose Hours, Date(s), Tag, Status(s), Saved Filters, and a Show jobs with errors switch | Workflow, Job, Agent |
| Workflow View | Choose Hours, Date(s), Status(s), Saved Filters | Workflow |
The Job View's columns are Date, Workflow, Job, Status, Return Code, Job Type, Start, End, Duration, Agent and Priority. A running job's Duration counts up every second; earlier builds moved it only when the list refreshed, about every 30 seconds. The Workflow View's are Date, Workflow, Status, Start and End. Every column except Status truncates with a tooltip carrying the full value, and the filter bar stays on one line, its controls shrinking as the window narrows rather than wrapping.
Status(s) and Tag report their selection as a one-line summary rather than listing every choice. Saved Filters carries a funnel icon, and the chip naming the applied saved filter sits under the button so selecting it never moves.
Job Type no longer has a header filter on the Job View, and Instance Name no longer filters the Workflow View. A saved filter or bookmarked URL that carried either one still loads — the retired value is ignored rather than narrowing the list invisibly.
A network firewall in front of the Continuum platform can block a request whose filter or search
value ends in a file extension, such as a job named Load.ini. The list then reads "A network
firewall blocked this request before it reached the server. This can happen when a search or filter
value ends in a file extension, such as .ini or .log. If your search ends in an extension, remove it
and search again." A * in place of the extension, as in Load*, still finds the job. Earlier builds showed
HTTP 403 and sent the blocked request again several times.
Changing page, sort, search, or a filter clears the current row selection (so a bulk action never acts on rows scrolled out of view).
Filtering by date and time
Date(s) takes more than one day. It opens on today, and its menu offers three ways to choose, plus a time window:
| Mode | What you get |
|---|---|
| A single day | That schedule date |
| Date Range | Every date from one day to another |
| Select individual Days | Specific, non-contiguous days — select a day, and add or remove more with Ctrl/Cmd and select |
| Time Range | A start and end time applied to each selected day |
You can type into the field instead of using the calendar — 05/29/2026, May 29, 2026, a range
with - or to, or a list of days. Text that is not a valid date stays in the field with an error
rather than being committed.
A range you leave half-picked is not kept. Choosing the first end of a Date Range filters to that one day straight away. If you then close the calendar without choosing the other end, the filter goes back to the dates, and any time range, you had before. Earlier builds kept the one day.
A Time Range narrows to jobs and workflows that started inside that window on each selected
day, in the scheduler's time zone, the zone the
times on this page are shown in. An end earlier than the start runs past midnight into the next day. A
missing start means 00:00:00; a missing end means 23:59:59. Rows that have not started yet never
match a time range.
Two limits worth knowing:
- A time range covers at most 31 days. Beyond that the field shows an error and the time range is not applied — the dates still are.
- Time Range and the Choose Hours preset exclude each other. Choosing a preset clears the days and the time range; changing the date field clears the preset.
Reordering the columns
Drag a column's label to move it, or reach it with the keyboard: tab to the label, pick it up with Space or Enter, move it with the Left and Right arrows, drop it with Space or Enter, or abandon the move with Escape. The sort control and the header's own filter box never start a drag, so selecting either still does what it did.
The order is remembered in this browser between sessions, per view — it is a personal preference, not part of the URL, so sharing a link does not carry it. Reset Columns, under Schedule Workflow, puts the visible view's column order and its column widths back to their defaults, and is unavailable while that view has neither been reordered nor resized. Resetting also opts the view back into any future change to the defaults.
The expand, selection and row-action columns are not reorderable: they stay first and last.
Columns can be resized too, by dragging the edge of a column header, and each view keeps its own widths in this browser. A column keeps its width when you move it. See Resizing table columns.
Name filters take a * wildcard
The name filters in the column headers — Workflow, Job and Agent — accept * as a
wildcard, and each carries a hover hint saying so. Where you put it decides what the
filter means:
| What you type | What it matches |
|---|---|
payroll | Any name containing payroll — the behaviour with no *, unchanged |
payroll* | Names starting with payroll |
*payroll | Names ending with payroll |
pay*roll | Names starting with pay and ending with roll |
*pay*roll* | pay somewhere, then roll somewhere after it |
Matching ignores case throughout. A value with no * in it still matches anywhere in the name,
so nothing you have saved changes meaning.
? is literal — it matches a question mark, not a single character.
_ and % are now literal too. They used to leak through as wildcards of the underlying query:
a filter of JOB_01 quietly matched JOB-01 and JOBX01 as well, and a single % matched
everything. Both are taken as the characters you typed, so a name that really contains one is now
findable.
A name filter is capped at 1024 characters — well above the longest name any object can have, so an over-long value from a saved report filter just matches nothing rather than failing. Beyond that, or a value that isn't a single piece of text, is refused.
Filtering by tag
The Job View's Tag filter takes several tags at once, and a job matches if it carries any of them. The list offers the tags actually present on jobs in the dates you have selected, with a search box, so you pick from what is there rather than typing from memory.
The tags offered span at most a year of schedule dates. When the date range has a start and no end — as it does when you open the Job View from a dashboard On Hold, Waiting or Running tile — the list offers the tags from roughly six months either side of today, or from the start date onward when that is more recent. A range with an end and no start offers the tags from the year before that end, or before six months from today when the end is later than that. Earlier builds offered no tags at all when you arrived from a dashboard tile.
You can also filter on a pattern: type one using * or ? and press Enter to add it as an
entry alongside the exact tags. A saved filter or URL carrying a single tag pattern from an earlier
build loads as one such entry.
The exact-name parameters a deep link can carry are unaffected. The Job History report's Workflow, Job and Agent name filters reach the same
matching, so * works there too — see Reports.
Clearing filters
Two actions reset the view's filters back to their defaults rather than leaving stale criteria applied:
- Clearing or deselecting a saved filter. Selecting a saved filter applies its criteria; clearing it puts the filters back to default instead of leaving the saved filter's criteria behind with nothing indicating they're still in force.
- View in Job View. Following a workflow into the Job View clears whatever filters were active, so you see that workflow's jobs rather than an intersection with a filter set for something else.
Moving between the two views
Switching views with the tab strip carries the date across. Whatever the view you leave is filtered to — its Date(s) selection, with any Time Range, or a Choose Hours preset — is applied to the view you open, so both views show the same days. Name, Status(s) and Tag filters stay with their own view. Two limits:
- The Workflow View always has a date, so a Job View with no date selected hands over as today.
- A Choose Hours preset is carried by name, and its window is worked out again when it is applied. Last 1 hour chosen at 09:00 in one view and carried across at 09:30 covers 08:30–09:30 in the other.
Earlier builds kept a separate date in each view, so switching views could show a different day.
The two actions below are different: each sets its own filters, including the date, from the row you came from.
Both directions are a push: the view you land on is filtered down to what you came from, and the one row you asked about is opened.
| From | Action | What the view you land on does |
|---|---|---|
| A job's detail | Workflow → | Switches to the Workflow View, filters to that workflow's name and schedule date (clearing Choose Hours, Status(s) and any saved filter), and opens that instance's row. Sibling instances on that date stay listed. |
| A job node on the workflow diagram | Node menu → View in Job View | Switches to the Job View, seeds the Job and Workflow name filters and the Date(s) filter from that job, resets to page 1, and opens that job's own row. |
The job node push identifies the job, not just its name. A job name is unique only within a
workflow instance, and the Job name filter matches anywhere in a name unless
you anchor it with a * — so the same name can belong to jobs in other workflows, to a
multi-instance job's fan-out, or to a longer name the search also matches.
Earlier builds seeded the job-name search and nothing else, which landed you on a list of rows with
nothing saying which one you had clicked. The workflow and date filters narrow it, and the row being
open is what identifies the job when they can't — two runs of one workflow on one date have no filter
that separates them.
The seeded date now lands on the Date(s) control's own range, so you can see it and change it. Earlier builds set a date axis the control doesn't render, which filtered the list invisibly.
The seeded Job name is typed into the filter rather than forced, so changing it — or clearing it to widen the list — abandons the rest of the push, including the row that was about to open. That is deliberate: the filter bar keeps telling the truth about what is applied.
A deep link into the Job View — the dashboard's failed-jobs widget, or a URL you saved — carries only what its own address says, so it seeds no workflow and opens no row.
Job Detail
Job detail
Selecting a job opens the job detail panel — a read-only view (the inline edit affordance has been removed). The hover overlay on a diagram node is a separate, lighter tooltip.
In the Job view the panel is a card beside the diagram, taking half the expanded row's width and its full height. The header and the tab strip stay put while each tab scrolls on its own, and a long job name wraps instead of being cut off. The Workflow view's own info panel is unchanged.
The panel's tabs
| Tab | What it holds |
|---|---|
| Output | What the job printed — see Output tab. |
| Summary | What happened on this run — Execution, Submission, Time Constraints, Completion, an SAP R/3 job's Child Processes, a legacy job's Agent Feedback, Kill Request (when the run was killed and by whom; it is cleared when the job next starts), a recurring job's recurrence details, and Retry — with Identity and Audit alongside. |
| Configuration | What the job will actually run with — see Configuration tab. |
| Status History | Every status change this job has been through. |
| Status Update | The actions you can take on the job, Restart among them. |
The Summary tab's Agent row shows the agent or pool the job is assigned to until it runs. Once
the job runs, the row names the agent that ran it. A held or waiting job, or one on a pool that has
not run yet, shows its assignment, by the same rule as the Agent column on the list. Earlier
builds showed no Agent row for such a job. A recurring job's Maximum Runs row shows the
limit the run enforces: a job with no Maximum Runs and no latest runtime reads 1 (see
Changing Success Reruns on a running job).
Status History is a tab, not a dialog. It used to open from an icon in the panel header; that
icon and the dialog are gone. Paging, refresh and the empty and error states are the same as the
dialog's, and each entry is a stack of labelled rows — timestamp, event, the old and new status, who
or what made the change, and any detail. A change made by a $JOB or $SCHEDULE event a person sent, through a Self Service button or the events API, names that person; one raised by a job, a workflow, a notification or an agent reads System. An entry's details open in a tooltip that wraps long text and scrolls
when it is tall, so a long reason — a container's build failure, for instance — can be read in full. History is fetched only while the tab is open, so
opening a job no longer loads it unasked.
There is no Restart button in the panel footer. The footer is gone; restart is in the Status Update tab with the other actions.
Earlier runs of the same job
The Status History tab covers the run in front of you. To look further back, use the Job Execution History button on that tab — it is there in every state of the tab, including while the history is loading and when there is none.
The dialog lists runs of this job across dates, which is the question Status History cannot answer: is this slow every night, or only tonight?
| Part of the dialog | What it shows |
|---|---|
| Job Name | Fixed to the job you opened it from, and read-only. The dialog does not search other jobs. |
| Date(s) | A single date or a range of up to 366 days. It opens on the job's own schedule date, so the first thing you see is the run you came from. A longer range shows Choose a date range of 366 days or less. on the picker, and nothing loads until you shorten it. When you pick a range on the calendar, nothing loads until you have chosen both ends, so the calendar stays where it is between the two picks. Close the calendar after choosing only the first end and the field goes back to the dates you had. A date you type is searched as that one day. |
| Run-time chart | Above the table: each run's run time against when it started, for the dates you chose — see below. |
| Columns | Date, Start Time, End Time, Runtime, Workflow, Job, Agent, Termination Status, Exit Code — all sortable, and paged. The first load shows placeholder rows. After that, while a new page, sort or range loads, the previous rows stay on screen dimmed until the new ones arrive, so the dialog does not change size or move under you, and the old rows never look current. |
The run-time chart answers is this job getting slower? at a glance. Each point is one run: its start time along the bottom, its run time up the side. Pointing at a point shows that run's start and run time.
- The run-time axis spans the lowest to the highest run time in the range, not zero upwards, so a small change stands out. It reads in hours, minutes and seconds — and milliseconds for a job that finishes in under a second.
- It covers the same runs as the table, for the dates you chose. Paging and sorting the table do not change it; changing the dates does.
- Above 5,000 runs it is drawn from a sample taken evenly across the range, always including the newest run, and says how many of the range's runs it shows, sampled evenly across the range.
- A range with no runs shows No runs in this date range. If the run times cannot be loaded you get the reason and a Retry; a chart already drawn stays on screen if a later refresh fails.
Choosing a row — or Job History Details on its row menu — opens Job History Detail for that one run, with two tabs.
Details lists the run's recorded values: Property Values, Agent Message, Workflow Name, Workflow Path, Job Name, a row per Tag, and Termination Description. A field with no value is left out rather than shown empty, so the table is only as long as the run has detail. Below the table, a legacy job's run that recorded any Agent Feedback shows it, grouped and labelled as on the job's Summary tab.
Property Values reads as name=value pairs, separated by |. An encrypted property is listed
by name with nothing after the =: its stored value is withheld — see Properties.
Notes is free text you keep against the job, for the next person who looks at it — why a run failed, what you did about it, what to check next time.
- Notes belong to the job instance, not to the single run you opened. Every run of that instance on its schedule date — a restart, a failure retry, a recurring run — shows the same notes. The same job in another workflow instance on that date has notes of its own.
- Anyone whose roles grant Job Instances Edit for the job can add, edit and delete them — reading them needs only Job Instances View — and each note shows who wrote it and when, plus who last edited it. Saving a note you did not change sends nothing, so it is not marked as edited. There is no ownership — if two people save an edit, the later save wins. If you edit or delete a note someone else has already deleted, the list refreshes to show the notes as they are.
- Deleting a note asks first, and then it is gone. There is no recovery.
- A note holds up to 4,000 characters, and a job keeps up to 500 notes. The next one is refused with A job can have at most 500 notes. Delete one first.
- Text typed and not yet saved survives a switch to Details and back, so losing your place does not lose your words.
A failed save keeps your text and shows the message the service returned, so you can correct it and save again rather than retyping.
Summary and Configuration lay out in two columns, and both drop to one when the card itself is narrow — which depends on the card, not on your window, so a wide screen with a narrow card still gets the single-column layout.
Summary → Completion carries an Arrived File row for a File Arrival job, giving the path the run actually matched rather than the pattern that found it. It appears only when the run reported a file, and it describes this run — a restart clears it.
Summary → Completion carries an Agent Message row
A legacy LSAM job that fails can send back a reason of its own — why the agent refused or could
not run the job. For a Windows, UNIX or SQL LSAM job that also returned an exit code, the
Termination row holds that exit code, so the agent's own words had nowhere to go and were
invisible. They now appear in an Agent Message row, directly after Termination. An IBM i
LSAM job works the same way with its message ID — SMA0037, say — in Termination: see
The message ID is the job's Exit Description.
- It appears only when there is something to show. The row is absent unless the agent sent a message and the exit code — or an IBM i job's message ID — displaced it from Termination. So the reason is visible exactly once, never in two places: an IBM i failure text that is the message ID itself is not repeated here.
- A failed SAP job carries an explanation. On an SAP R/3 Job or an SAP BW Process Chain that fails,
the row says what the exit code means — The process chain did not finish successfully., say —
and adds the agent's own text after a
|when that says more than the bare code. See What the job reports for SAP R/3 and for SAP BW. - It is per-run evidence. Like Arrived File, it is cleared when the job is re-armed, restarted, or dispatched again — it describes this run.
- A failure the relay itself reported is not an agent message, and is never put here. The generic Job failed and the relay's own reasons — a job lost after a reconnect, a runtime ceiling exceeded, a kill that timed out — carry no agent detail.
A UNIX LSAM echoes the command line and its arguments back in its failure reason when it cannot start the job. If that command line was built from an encrypted property — or from a credential the job type supplies — printing the agent's text would make a secret readable again, which is exactly what encrypting it promised to prevent.
So when the run's command line carried a secret, the row reads Agent message withheld — this run's command line carried a secret instead of the agent's text. The same substitution is applied to the termination fallback and the exit-criteria trace.
It fails closed: if it cannot be established that the run carried no secret, the message is withheld. A job whose command line holds no encrypted value is unaffected.
The Output tab now shows the server's reason too. When output cannot be fetched, the tab reports the error the server actually sent rather than a bare status code — so, for instance, the gateway failure that explains why an IBM i job has no output is readable. A job with no LSAM job number is also described correctly now: the message no longer claims output is not available for this job type, which was wrong in its commonest case — an undispatched or restarted legacy run.
The panel keeps its Date and Workflow after an action. Editing a job, or restarting it, used to blank those two fields until you hard-refreshed or navigated away and back. Every job action — edit, restart, hold, release, start, kill, cancel and the rest — now returns the job with its date and workflow intact, so the panel stays populated. (Restart still clears the exit code and agent fields; that is the reset it is meant to perform.)
A job that overran its Max Run Time
A job that has run past its Max Run Time is marked wherever you look at it, in a warning tone rather than an error one — the job has not failed:
- Beside the job's status — in the Status column of the jobs list, and beside the Status value at the top of the detail panel's Summary tab. It is a separate badge from the status badge, and the status is unaffected: the job reads Job Running while it runs and then whatever it earns when it finishes.
- On the workflow diagram, as a timer icon beside the job card's status icon, on the expanded row in both Processes views. Hover it for Exceeded Max Run Time. On a Job View context card that is not the job in focus the icon is greyed, like the rest of the card.
- On the Summary tab, in the Time Constraints section, as a Max Run Time Exceeded row giving the time the overrun was detected. The limit itself stays on the Configuration tab and is not repeated here.
An overrun latches without the job's status changing, so the diagram polls for it while it is open — an already-open diagram picks the icon up rather than waiting for something else about the job to move.
The icon is hidden while you scrub the timeline. Playback reconstructs each job's status as of the point you have scrubbed to, but the overrun marker is a live value with no history behind it, so showing it against a reconstructed status would date it wrongly. The same applies to a job's termination description.
Time Constraints appears for a job whose only timing value is the overrun, which is the common case — Max Run Time is configured independently of every other timing field.
The badge stays after the job completes. A job that overran and then finished OK is exactly the case the status alone cannot show, which is why the marker outlives the run rather than being a live-only indicator.
There is no filter for it — the jobs list filters on status, and this is not a status. See Exceeding Max Run Time for what the platform does about the overrun, and what it deliberately does not do.
Configuration tab
The Configuration tab describes what this job instance will actually run with, not what the workflow's design says:
| Section | Read from |
|---|---|
| Task Definition | The job instance. |
| Agent Assignment | The job instance. |
| Frequency Runtime Info | The parent workflow's build snapshot. |
Task Definition and Agent Assignment read from the instance, so they reflect anything that changed between the design and this run — a deploy-time transform rule that retargeted the agent, or an edit made to the instance itself. Earlier builds read both from the workflow's frozen design snapshot, so a job that had been retargeted on deploy showed the agent it would have used rather than the one it did.
Both sections render even when the job isn't present in the build snapshot at all. Frequency Runtime Info still comes from the snapshot, so it's the one section that can be unavailable for an instance the snapshot doesn't cover.
Output
Output tab
The Output tab shows what the job printed. Its header always carries the job's Date, Exit Code and Workflow as a row of labelled values, so a log read out of context still identifies its run, and the tab's actions are grouped together on the right of that header. The log or the file list fills the rest of the card and scrolls inside it, so the header and the actions stay put however long the output is.
The tab has two shapes, chosen by whether the run has an LSAM job number — which a legacy (LSAM) run gets shortly after its agent accepts the job, while the job is still running:
| The run | What the tab shows |
|---|---|
| Has an LSAM job number (a dispatched legacy run) | A list of output files the agent holds, any one of which opens in place. |
| Has none (a Continuum-agent job, a container job, or a legacy run not yet dispatched or just restarted) | The output returned with the job, as one log, with Refresh output. With nothing returned yet the tab reports No output. |
| A job with no job type recorded | No output, and no refresh or download control. |
Open shows the file you are looking at in a new browser tab, at that tab's full width. It is the way to read a long log: the in-panel view materialises only the first 500 KB of a file and says so — Showing the first 500 KB of a 2,400 KB file. Use Open to read all of it. Open and Download always work from the whole file, not the truncated view. If the browser blocks the tab, the panel says so rather than doing nothing.
Download never truncates either. All downloads everything, and the caret beside it offers the parts by name — every output file on a legacy job, and the single log on a Continuum-agent job.
Both shapes also carry a refresh: Refresh from agent on a legacy job, which is described below, and Refresh output on a Continuum-agent job, which re-reads the job.
Output files from a legacy agent
A legacy agent keeps its output on the machine, so OpCon Continuum has to ask for it. The first time you open the tab for a job, it fetches the file list from the agent through the relay and stores it; after that the stored copy is served, which is why the tab opens instantly on a second visit. The Output Files heading carries the file count and which of the two you are looking at. The same applies to each file's contents, fetched and stored the first time you open it.
Refresh from agent discards the stored copy and asks again.
While the job is running, nothing is stored. The tab lists a running legacy job's files too — except an SAP R/3 Job's or an SAP BW Process Chain's, which can be listed only once it has ended — but every look asks the agent, so you always see the output as it stands — and a list read while the job was still writing is never kept as its final one. Once the job stops running, the tab fetches the list again and stores it as above. A job marked Failed or Finished OK while its process is still running on the agent is stored as usual; when that process does end, the stored copy is discarded, so the next look reads the finished output.
The files are listed in the order the agent sent them, so a z/OS job's SYSOUT reads JESMSGLG,
JESJCL, JESYSMSG as the mainframe lists it. A listing stored before this ordering was kept is
shown in name order. If the agent lists the same name twice, the first is kept.
A z/OS job's output, of any z/OS job type, is shown as it is laid out on the mainframe: each row keeps its spacing, and a file's contents scroll sideways rather than wrap. It can be listed only when the z/OS agent reports where it filed it, and the tab says so in plain words when it has not — see Its output.
An MCP job's output can be listed while it runs as well as once it has ended: where the agent did not report where it filed it, Continuum works it out from the job's MCP usercode and the mix number in its status description — see Its output. An OS 2200 job's output, like a z/OS job's, can be listed only when its agent reports where it filed it — see Its output.
An agent with no JORS Port cannot list output. The tab says JORS port not configured for agent: …, naming the agent; set the port on the agent — see Agents and agent pools.
For a running IBM i job, the request needs a job identifier the agent reports a moment after the job starts. Asked before then, the request is refused with a reason saying so — the job is still running and its agent has not yet reported that identifier. Try again a few seconds later.
Two limits are worth knowing:
- Only files the agent listed can be read. Reading a file the list does not hold is refused, so the tab cannot be used to reach arbitrary paths on the machine.
- Output is per environment. A job that belongs to another environment reports not-found for its output files, the same as for every other read — see Environments.
Re-running a job discards its stored output listing. The listing and the file contents belong to a run, while the stored copy hangs off the job instance — and a restart, a failure retry, a recurring restart and a force start all re-use that same instance. Each of them now clears the stored copy as the job is dispatched, so the Output tab reads the new run's own output instead of serving the previous run's files as though they were the current ones. You do not have to remember to Refresh from agent after a restart.
Three things are dropped together, because each of them was independently capable of serving the previous run's output as the current one:
- the stored file listing and file contents;
- the identifier the previous run was known by on the legacy agent. A UNIX agent names an output file after that identifier, so a request still carrying the previous run's one matched the previous run's file exactly. A listing that comes back late — the request went out before the job was re-dispatched — is now checked against the identifier it was sent with, and discarded rather than cached, so a slow answer from the old run cannot become the new run's output;
- the inline output a non-legacy job returns with its completion, which is likewise written only when a run finishes.
The page's own copy is held per run as well. Clearing the stored copy on the server fixed only half of it: the browser kept what it had already read under the job, so re-opening the tab after a rerun could redisplay the previous run's listing and file contents without asking the server anything. What it holds is now tied to the run's identity on the agent — the legacy job number, and for IBM i the request parameters its agent reports a moment later — both of which are cleared at dispatch and reissued for each run. A rerun is a different run, so the tab fetches it.
There is no automatic refresh, and the one that used to run was removed because it did damage rather than good: marking a Failed job Finished OK restamps its end time, which made a job that had finished days earlier look as though it had just ended, and the next visit to the tab issued a refresh nobody asked for — purging the stored output, or hanging for up to a minute on an agent that was offline. Use Refresh from agent when you want the agent asked again.
A restart that is never dispatched is cleared too. Restarting a job on hold, or a restart whose dispatch fails immediately with an initialization error, used to leave all of the above in place — indefinitely, in the second case — so the Output tab showed the previous run's output under a job that had not run. The clear now happens on the restart itself as well as at dispatch.
Retrieving output from a UNIX or IBM i agent depends on how the request names the job, and that naming is built by the relay's connector. Two things went wrong on older relays:
- A UNIX request matched nothing — and because the UNIX agent falls back to its own global log,
the tab came back populated with the agent's
logfileanderrfileinstead of the job's own output, reporting success the whole way. It looked right and was wrong. The platform now discards files that do not belong to the job before it stores a UNIX listing, whichever relay answered, so on an older relay the tab comes back empty rather than wrong. - An IBM i request returned nothing at all. IBM i does not name a job the way the other platforms do: the request has to carry the OS/400 qualified job name — job name, job user and the job number assigned when the job was submitted — and that number exists nowhere in OpCon Continuum. It is reported by the agent itself, and the connector was not reading the part of the agent's status message that carries it.
Current builds send the naming each platform expects, and refuse a request that cannot match rather than sending it — so a failure is visible and retryable. IBM i output is now retrievable rather than refused. If UNIX or IBM i output still comes back wrong or empty, check the relay build first: the relay has to carry these changes for them to take effect, and nothing in the cloud can yet tell an out-of-date relay from a job that genuinely produced no output. Windows and SQL agents are unaffected and were always correct.
If a job's output already looks like the agent's global log rather than its own, use Refresh from agent after the relay is upgraded — the wrong list was stored, and only a refresh replaces it.
Permissions
Who can do what on Processes
Every read and every action here is checked against your roles, whether it is taken on the page or through the API. Opening Processes needs the Workflow Instances View permission.
- Reading needs a permission per list, in the selected environment: Workflow Instances View for Workflow View and Job Instances View for Job View. Each list, and the run's counts, show only the instances whose workflow is in a workspace where your role grants that permission. Job Instances View on its own does not open the page.
- Opening one instance needs its view permission in that instance's workspace. Without it, the instance answers as not found, as if it did not exist.
- An action on one workflow or job instance needs its permission in the workspace of the workflow the instance was built from: Job Instances Execute for every change of a job's status (Hold, Release, Skip, the marks, Cancel, Kill, Restart and Force Start); Job Instances Edit for editing a job and its notes; Job Instances Delete for Delete; Job Instances Create for Add Job; Workflow Instances Create to schedule a workflow; Workflow Instances Execute for Start, Hold, Release, Close and Cancel; Workflow Instances Edit for properties; and Workflow Instances Delete for Delete.
- A workflow that has since been deleted keeps its workspace, so its remaining instances are read and acted on under that workspace like any other.
- A refused action changes nothing. If the role or workflow lookup is briefly unavailable, the action is refused rather than allowed, and can be tried again.
The full table is in Running workflows and jobs are checked per workspace.