[
https://issues.apache.org/jira/browse/CAMEL-26699?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Claus Ibsen reassigned CAMEL-26699:
-----------------------------------
Assignee: Claus Ibsen
> camel-jbang - chaos screen: findings of a hands-on round (poll targets,
> expectation count, pick order, effect of delays and polls)
> ----------------------------------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-26699
> URL: https://issues.apache.org/jira/browse/CAMEL-26699
> Project: Camel
> Issue Type: Improvement
> Components: camel-jbang
> Reporter: Claus Ibsen
> Assignee: Claus Ibsen
> Priority: Major
> Fix For: 4.24.0
>
>
> Findings from a hands-on round of the chaos screen (CAMEL-26695) with a small
> demo shop: an orders timer, an HTTP stock and payment call behind a circuit
> breaker, a seda shipping route, a file poller, a dead letter channel with 2
> redeliveries.
> h3. Wrong or misleading
> # *Polls fail is offered for a timer.* The consumer status marks both the
> timer and the file consumer {{scheduled: true}} ({{polling: false}}), so the
> add form offers {{timer://orders}} for "Polls fail" / "Polls are skipped".
> {{at=poll}} hooks only a scheduled poll consumer: the rule
> {{endpoint=timer://orders error at=poll}} matched 0 times, and was added
> without a warning. The form should offer only poll consumers, and the rule
> check should warn when {{at=poll}} matches no poll consumer.
> # *The count of a decided expectation stops.* {{expect=orders:handled<=0}}
> went unmet at the first exchange handled by the error handler, and its count
> stayed at 1 while the rule's effect and the route counts said 6. That the
> outcome no longer changes is by design; the count shown should go on (or say
> "unmet at 1, now 6").
> h3. Picking and reading
> # *Endpoint pick order.* For "Is slow" the first endpoint offered is
> {{seda://dlq}} (the dead letter channel), then {{direct:}} and {{seda:}}
> internals; the HTTP call is 6 of 8. Offer the calls to other systems first
> and the dead letter channel last (it has its own recipe).
> # *Column widths.* The EFFECT column is cut ({{62↩ 1…}}) while TARGET has
> room to spare on a wide terminal; FAULT cuts a lifecycle rule to {{lifecycle
> action=stop duration=5}}.
> # *A delay shows no impact.* A delay rule's effect is "43 completed", while
> the route's p99 went from 1 ms to 907 ms and the throughput from 2.0/s to
> 1.7/s (seen only on the Routes tab). Show the latency the rule added, or the
> route's time next to it.
> # *A poll fault shows no effect.* For a poll rule the EFFECT is empty; it
> could say how many polls failed or were skipped.
> # *Recent faults repeat an exchange* once per redelivery attempt (the same id
> three times); group them ("…286B ×3 → dead letter channel").
> h3. Driven by an AI (the TUI's MCP)
> # *A confirm dialog drops Esc too.* Keys from MCP are dropped while a confirm
> is open, by design, but that includes Esc: an AI that opened the add confirm
> cannot close it, and it stays until a person presses Esc. Let an MCP client
> cancel (never confirm).
> # *{{!}} after {{tui_navigate}} to a Diagram node* takes the route entry
> instead of the node: the selected box follows the arrow keys only.
> The redelivery counters that the Routes tab lacks are CAMEL-26698.
> _Claude Code on behalf of Claus Ibsen_
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
