Decide in Advance What Happens When a Transfer Goes Wrong

Set what INNORIX should do for low storage, a changed source, a name clash or an overlapping schedule, then read what the policy actually did.

IT Engineers

What this solves#

A transfer is stuck at three in the morning. Someone has to decide.

  • The target disk is full. Wait, or skip this device?
  • The source file changed mid-transfer. Send it again, or skip it?
  • Yesterday's run has not finished and today's scheduled time has arrived. Overlap, or skip this one?

Error handling is where you write those decisions down in advance. Once written, nobody has to wake up: INNORIX applies them and records what it did and why.

And nine of them you never configure at all. INNORIX always handles those itself.

Who this is for#

Role When
Operations Transfers run unattended at night and over weekends
System administrator So many devices that a few always have a problem
Security and audit The handling has to match policy

Before you start#

Where to set it#

  1. Flows → pick the Flow → Edit flow.
  2. Press the Assurance badge in the toolbar.
  3. Open Error handling in the lower half of the dialog. The summary reads INNORIX recommended while everything matches INNORIX's own values, and N custom once anything differs.
  4. Each group lists issues, and each issue takes an action.
  5. Press Apply, close the dialog, then Save & activate.

Reset to defaults, in the Success conditions row above, returns the actions set here as well, not just the success conditions. What it returns to is this workspace's defaults. With nothing set at the workspace level, that is INNORIX's recommended values. Cancel puts back whatever was there when the dialog opened.

What you can set#

The "Default" column below is INNORIX's recommended value. Leaving them alone is a valid choice.

Read this first. What a Flow actually starts from is this workspace's Assurance defaults. If the workspace changed an action, your Flow starts with that value, not the one in the table below.

You can tell which is which.

What you see Meaning
A Default badge on the row That row is INNORIX's recommended value
A Workspace badge on the row That row came from the workspace defaults
Workspace defaults on the badge beside the flow name This Flow is using the workspace defaults as they are
Error handling reads N custom although you changed nothing Expected. That summary compares against INNORIX's own values, so anything the workspace set counts as different

The workspace defaults live under workspace settings, in Assurance defaults. Anyone can read them; only the workspace owner or an administrator can change them.

Network#

Issue Default Meaning
Insufficient storage Keep retrying The target disk is full. Keep trying until space appears
Recovery not progressing Keep retrying Recovery is not making headway. Keep trying

Source#

Issue Default Meaning
Source changed Transfer the file again The source changed mid-transfer. Send the new content
Source deleted Skip the file The source is gone. Skip just that file
Source file in use Skip the file Another program holds it. Skip just that file

Destination#

Issue Default Meaning
Destination file exists Keep both A file of that name is there. The dialog notes "Set in each step's save options.", so the real rule lives in the step
Destination file in use Skip the file Another program holds the target file
File system naming conflict Skip the file The file system cannot accept that name

Schedule and time#

Issue Default Meaning
Schedule overlap Skip this run The previous run has not finished. Skip this occurrence
Deadline missed Continue The deadline passed. The transfer continues, but the contract is violated

The actions#

Action What happens When
Keep retrying Keep trying until it works Temporary problems: full disk, a brief lock
Retry later Try again later, holding other files meanwhile. If it still cannot, that file is skipped You know it will clear soon
Transfer the file again Send that file from the start The source changed
Skip the file Leave that file out and carry on One file must not stop everything
Skip this device Leave that device out and carry on One device must not block the rollout
Skip this run Skip the whole occurrence Schedules overlapped
Queue one Queue the occurrence instead of skipping it. It runs once the previous one finishes The occurrence must not be dropped
Continue Record it and carry on Worth knowing, not worth stopping for
Stop the run Halt the run Continuing would be unsafe
Request review Wait for a person A decision that must not be automatic
Keep both Keep both files of the same name Overwriting is not acceptable
Overwrite Overwrite the file of the same name Only the latest copy matters. Offered on Destination file exists alone

Each issue offers only some of these. The table above is all twelve actions there are; opening one issue's list in the dialog shows two to four of them. An action that is not listed cannot be used for that issue.

Issue What you can choose
Insufficient storage Keep retrying / Skip this device / Stop the run
Recovery not progressing Keep retrying / Request review / Skip this device / Stop the run
Source changed Transfer the file again / Skip the file / Stop the run
Source deleted Skip the file / Stop the run
Source file in use Retry later / Skip the file
Destination file in use Retry later / Skip the file
File system naming conflict Skip the file / Stop the run
Schedule overlap Skip this run / Queue one
Deadline missed Continue / Stop the run

Destination file exists is not chosen here. Its value (Keep both / Skip the file / Overwrite) comes from the step's save options, and the dialog only reports it.

An issue set to Request review shows up under Runs requiring attention as Pending approval on the Assurance screen. The card reads N devices are waiting for your review, and retries on those devices are paused until someone decides ("Retries are paused for these devices until someone decides.").

View report puts a review box at the top with three choices.

Choice What it means
Resume You looked, carry on as before
Skip this device Leave that device out, the rest continue
Stop the flow run Halt the run

Choosing one clears the box and leaves a record in Activity naming the choice and who made it, as in Review resolved: resumed by …. While it waits, Contract is Pending and the condition rows read Judged when the run finishes.

This is not an approval before the transfer starts. Pre-run approval is a Portal feature.

When your choice does not take effect#

Picking an action is not the same as the action happening. Four things can stand between the two, and each of them shows on screen. Check them before concluding that a setting is broken.

What you see What it means What to do
Applied beside the issue The server enforces this one. Your choice is live Nothing
Coming in a later update beside the issue That issue is enforced by the agent, and the agent cannot do it yet. Your choice does nothing for now Treat the issue as unhandled until the tag goes
An Agent tag beside the issue Enforced by the agent, with "Devices with older agents use the default handling" Update the agents, or expect the old ones to behave differently from the new ones
"Choices are recorded but not applied yet." The whole policy is in observe mode. Decisions are written down, not carried out Nothing on this screen. The mode is set on the server side

Observe mode is the one that surprises people. A device that should have been skipped is not skipped; Activity records what the policy would have done and tags the decision Observe instead of Applied, reading something like "Policy would skip this device (observe mode)". The wording of that line comes from the server. Everything reads as if the policy ran, and nothing happened.

Schedule overlap carries no tag at all. Skipping the overlapping run is what the product already does, so there is nothing extra to enforce.

The nine you never configure#

The Handled by INNORIX group in the dialog, labelled "9 items, handled automatically". You cannot turn these off or change them.

Group Issue What INNORIX does
Network Connection interrupted Automatic reconnect and resume
Transfer process restarted Resume from the last confirmed point
Transfer stalled Detect and reconnect
Receiver backpressure Slow down and continue
Destination Integrity mismatch Transfer the file again; fail if it repeats
Security Malware detected Follows the security integration setting
Encryption key mismatch Fail
Retry blocked by device policy Stop the run
Audit logging unavailable Record and continue; notify admins

Note that the four interruption cases live here. When a customer asks whether a transfer resumes after the line drops, this group is the answer. It works without being configured.

Worked examples#

Situation Change Why
Store terminals, some always powered off Insufficient storage → Skip this device One store must not hold a nationwide rollout all night
Production recipes, a wrong file is unacceptable Source changed → Stop the run Stop and let a person check
Nightly log collection, due next morning Schedule overlap → Skip this run (keep the default) Safer to skip today than to overlap
Financial evidence transfer Destination file exists → Keep both (keep the default) Overwriting would destroy evidence
Medical imaging Source file in use → Retry later A study still being written can be fetched when it closes

Reading the result#

When a policy fires it appears in the Assurance Report. A real example:

Signage update · 50 devices
Transfer Status   Success     49 / 50    1 skipped
Contract Status   Satisfied   All 4 conditions met

Conditions                                       All 4 conditions met
  ✓ Target completion      49 / 49 targets   1 skipped   ≥ 95%   Passed
  ✓ File accuracy          49 passed  MD5                        Passed
  ✓ Stored at destination  49 / 49                               Passed
  ✓ Security & evidence    no violations                         Passed

Devices   50 devices    Skipped (left out of the rate) 1    Completed 49
Issues & Recovery   1 recorded event · 3 recovery attempts · on 1 device
                    1 unresolved
Activity   8 events · 1 policy decision

The thing you must understand: a skipped device leaves the denominator#

There were 50 targets, yet Target completion reads 49 / 49. The skipped device was left out of the rate, which is why the 95% condition passed.

The two blocks above group the devices differently: this one by Skipped and Completed, the clean run below by On time. That follows the contract, not the outcome. With a deadline set the groups are Late and On time; without one they are Completed, Completed with issues and Not started. Skipped and Failed stand on their own either way. See Reading Per-Device Results and What Is Missing.

This is deliberate. The policy said to skip that device, so the judgement skips it too. But it can read as "49 of 50 received it and we scored 100%", so an operations report should quote the Skipped count from Devices alongside the rate.

Do not leave the skipped device alone. The Unresolved count under Issues & Recovery is the reminder. Fill it in with the right kind of resend.

"No issues" does not mean nothing happened#

Some records are notices, not issues. A same-name file saved as name(1), a filename normalised, a timestamp not carried over, a destination path that changed. The transfer handled each by its rule, so the server leaves them out of the result, out of the counts and out of Issues & Recovery.

A run that saved every colliding file under a numbered name therefore reads Recovery: Verified, Contract: Satisfied and Issues & Recovery: No issues. Nothing on that screen says a rename happened.

They are still in Activity. That is the only place to answer "did anything get renamed, or lose its timestamp". Twelve kinds are treated this way, including File name conflict (DST-03), Duplicate file delivery (DST-02), File name normalized (SRC-12) and Timestamp not preserved (DST-13).

This matters most when Keep both is the chosen action: it works, it is recorded nowhere you would notice, and the destination quietly accumulates numbered copies.

A clean run#

Devices   100 devices   On time 100
Issues & Recovery   No issues

No issues means no policy fired at all.

Errors and exceptions#

What you see Meaning What to do
Badge reads Defaults apply The settings could not be loaded The run uses recommended values. Check the server connection
Stuck at Pending approval An issue set to Request review occurred In the review box under View report, choose Resume, Skip this device or Stop the flow run
1 unresolved remains An issue was not recovered Find it in the report and handle that device separately
Deadline missed but the transfer continued INNORIX's recommended value is Continue Change it to Stop the run if it must halt
The same issue every day An environment problem, not a settings problem Check the device and its disk. Do not paper over it with an action

Operating notes#

  • Keep retrying conflicts with a deadline. Retrying past the deadline violates the contract. Where time matters, Skip this device is often the right answer.
  • Skipping a file and skipping a device end differently in the contract. A run that dropped a device for insufficient storage with Skip this device left that device out of the denominator and finished Satisfied. A run that dropped one file with Skip the file because the source changed broke Stored at destination and finished Violated. Both read as "skip", and the verdicts are opposite.
  • Retry later ends in a skip too. When the wait does not clear it, the file is left out, which lands where Skip the file lands. The policy decision in Activity records how many times it tried before giving up.
  • Skip this device changes the denominator, as above. It loosens a percentage condition, so pin the machines that matter under Required devices.
  • Use Stop the run sparingly. One device can halt a 1,200-device rollout.
  • Trust the defaults first. INNORIX recommended fits most environments. Change only what actually bites you in operation.
  • Changes apply from the next run.
  • Before changing an action, read the N policy decisions count on the Activity line (1 policy decision when there is one) to see how often it really fires.

By industry#

Industry How it is used
Manufacturing Source changed → Stop the run for recipes and machine settings. A wrong file reaching equipment is the real risk
Retail Insufficient storage → Skip this device for store terminals. The out-of-hours window is too short to wait
Finance Keep Keep both, since overwriting destroys evidence. The automatic admin notice on Audit logging unavailable is useful for audits
Healthcare Retry later for studies still being written. Integrity mismatch is automatic, so there is nothing to set
Public sector and defense Use Request review broadly so people, not rules, make the call
Broadcast Air time is absolute, so Deadline missed → Stop the run keeps late files off air