Troubleshooting Analyze

1. Introduction

This page is organised by symptom rather than by screen: find the thing you are seeing, and the cause and the fix follow. The entries are ordered roughly by how often they are hit, so the cheapest checks come first.

2. The results grid is empty

2.1. The date window excludes your messages

By far the most common cause. Every result screen opens with From set to yesterday at 00:00:00 and To set to the current time plus one hour. A message that arrived outside that window is not missing, it is simply outside the query.

Widen From and press Apply filters. The filter panel is described in Analyze.

2.2. You changed the filters but did not press Apply filters

On the Trace overview the filter fields do not drive the query on their own. Nothing is re-queried until you press Apply filters, so the panel can show one thing while the grid still shows the result of an older query.

The applied-filter chips above the grid are the reliable indicator: they describe what the agent was actually asked, not what the panel currently displays.

The Business entities overview works the other way round: changing the From or To date re-queries at once, without Apply filters.

2.3. You pressed Clear filters and expected new results

On the Trace overview, Clear filters empties the Interface, Flow, ComponentType, Sender and Service fields and removes the chips above the grid. It leaves the From and To dates as they were, and it does not re-run the search, so the grid keeps showing the previous result set until you press Apply filters.

On the Business entities overview, Clear filters re-queries immediately, so the grid does follow the cleared panel there.

2.4. The agent has no Analyze adapter installed

The symptom is every Analyze screen empty: no traces, no messages, no business entities, and no error at all. The agent is most likely answering every Analyze call with a 404 because the Analyze adapter is present but not installed.

The ConnectPlaza® Agent ships the adapter staged, ready to install, in .data/update/adapters/. Staged is not running, and restarting the agent does not promote it. The wiretap still stores its copy of each message, so the database grows, but nothing serves that data to the Studio and nothing evaluates a pipeline.

To confirm, open Deploy, select the agent and look at its System interfaces. The ConnectPlaza® Analyze Adapter should be listed among the installed interfaces and running. If it appears only under Ready to Install, that is the problem.

To fix it, install that system interface, as described in Analyze installation on a running ConnectPlaza® Agent. It starts by itself once installed.

3. The search box does not find my message

The box in the grid header filters only the rows already loaded on the page you are looking at. It is not a database search: rows on other pages, and rows outside the current date window, are never matched, however you spell the term.

To search the stored data, use the filter panel and press Apply filters.

4. Refresh did not do what I expected

There are two refresh actions and they do different things.

Control Where it is What it does

Refresh filters

In the filter panel

Reloads the options in the dropdowns from the agent: the interfaces, flows, senders and services it has seen. Use it when a newly deployed flow is missing from a dropdown.

Refresh results

In the grid toolbar

Re-runs the current query and reloads the rows. Use it to pick up messages that arrived since you last searched.

5. My business entity does not appear in the Overview

5.1. You have not selected both a Definition and a Version

The Overview adds one column per definition property only once both a Definition and its Version are selected. Until then there are no property columns, and the Property filters section has nothing to offer in its Column dropdown.

5.2. The message never reached the pipeline

Work back along the chain described in Analyze concepts, in this order:

  1. Does the message appear under Traces, messages and comparing? If not, the wiretap is not storing it. Check that the interface is installed and running.

  2. Is the pipeline bound to the definition you are filtering on?

  3. Is the Agent Mapping rule bound to the correct wiretap point, and to a Snapshot or Release of that pipeline?

  4. Is that mapping assigned to this agent under Deployments?

5.3. Every rule returns nothing, on a message that clearly holds the data

If the message arrived over an HTTP Listener, check the listener’s Expected Content Type. Left at DEFAULT, the listener delivers the body as a byte array, which Analyze stores Base64 encoded. The stored content is then valid text, but it is no longer your JSON or XML, so every JSONPath, XPath and regular expression rule fails to match and no attribute is filled.

Set Expected Content Type to STRING on the listener and send a new message. Messages that are already stored keep the encoding they were stored with.

5.4. The pipeline produced an attribute called unknown

When an assignment produces no result, the entity is still created and the missing attribute is written as an empty TEXT attribute named unknown, which then appears as its own column in the Overview. Read it as a signal that a rule upstream of that assignment matched nothing.

6. I cannot save or promote a pipeline

6.1. Cannot save: some assignments do not match the selected Business Entity

Changing the bound definition leaves behind assignment nodes for properties the new definition does not have. Connected orphans are flagged as invalid on the canvas, and Save Draft, To Snapshot and To Release stay disabled until they are gone.

Delete the flagged assignment nodes and save again.

6.2. The definition I just created is not in the Definition dropdown

A Draft never leaves the Studio. A pipeline can only bind a Snapshot or a Release of a definition, so a definition you have only saved as a Draft does not appear in the list. Promote it with To snapshot first, as described in Analyze concepts.

The same rule applies one level up: an Agent Mapping rule can only target a Snapshot or a Release of a pipeline.

6.3. The canvas is empty on a pipeline I know has content

A Draft opened directly by its URL, in a cold browser tab, can render an empty canvas. Reload the page before doing anything else, and in particular before running a simulation: a simulation started on an empty canvas submits an empty pipeline and reports success with no entities.

6.4. Nothing on this screen can be edited

Only a Draft is editable. On a Snapshot or a Release the canvas is locked, the ports are disabled and the configuration fields are read only. Use To draft to copy the version you are looking at back into your personal draft.

To draft overwrites your current Draft, and so does releasing from a Snapshot, which additionally updates the definition reference. Check what is in your Draft before you do either.

7. The simulation did not do what I expected

7.1. The Select Message dropdown is empty

The simulation’s Select Message list is filled from the Compare Messages selection, not from anything on the pipeline screen. Until at least one message has been added there, the simulation has nothing to run against.

Go to Traces, messages and comparing, open a trace, tick the checkbox on the message you want, and then return to the SIMULATION tab.

7.2. The simulation reported no error, but produced nothing

A simulation whose pipeline fails still completes successfully as a request: the simulation ran, the pipeline is what failed. There is no error banner. Read the verdict chip on the Simulation Results dialog instead.

Chip Meaning

Success

Every rule ran and every assignment resolved.

Partial

Some assignments resolved and some did not. The entities that could be produced are listed, and the rest are explained on the ERRORS tab.

Failed

Nothing could be produced.

The ERRORS tab names the node that failed and gives the reason, so a broken rule can be found on the canvas instead of guessed at.

8. The agent is running old configuration

Promoting a pipeline or a mapping to Snapshot sends the new configuration to the agent, but the Studio does not wait for a confirmation and does not report a failure. A snapshot can therefore report success while the agent is unreachable and still running the previous content.

If an agent looks out of date, first check that it is connected, on the Agent Selection screen. The agent also pulls its configuration again about ten seconds after it reconnects, so restoring the connection is usually enough.

9. Analyze is filling up the database

Cleanup is disabled by default. Nothing is ever deleted until an administrator enables it. See Analyze properties for connectplaza.analyze.cleanup.enabled, which covers messages, and connectplaza.analyze.be.cleanup.enabled, which covers business entities. Each has its own schedule and its own expiration.

10. A stored message is not the whole message

Analyze caps a stored message at 1 MB in total. Above that, each part is trimmed to an equal share of the budget and its stored content is replaced by a marker naming the limit and quoting the first part of the content.

Message history is not capped. It is stored in chunks of at most 200 characters, spread over as many connectplaza-history-N.M properties as it takes, and the message id history is split the same way over connectplaza-messageid-history-N. Nothing is lost there.

This affects both what you can see in Message details and what a pipeline has to work with, so extract what you need early in the flow rather than wiretapping a large payload late.

11. The same thing has three different names

One dimension, meaning which wiretap point a message passed, is labelled differently depending on where you are looking.

Screen Label

Trace overview filters

ComponentType

Trace details grid

Component

Business entities overview filters

Wiretap

They are the same dimension, but they do not carry the same value. The ComponentType filter on the Trace overview and the component column in the Trace details both show the component’s type, while the Business entities overview’s Wiretap dimension carries the component’s name. Two wiretaps of the same type in one flow therefore look identical on the Trace overview and can be told apart on the Business entities overview.

The Trace overview has no column for this at all. Its steps column is a count of the wiretaps in the trace, not a type.