Thread Dumps and Diagnostics¶
When QIE is slow, stuck, or using more memory than expected, the answers come from the JVM rather than from the channel logs. Help -> Monitor Server Resources opens a dialog with four tools for getting them.
They differ enormously in cost. Two are free to run on a production system; one pauses processing for seconds; one pauses it for minutes. Reach for them in this order:
| Tool | Cost | Use it to find out |
|---|---|---|
| View Threads | None | What every thread is doing, right now. Start here. |
| Run GC | Negligible | Whether memory is genuinely retained or just uncollected. |
| Thread Dump | Pauses briefly | A snapshot of every thread's stack, as a file to attach to a ticket. |
| Heap Dump | Pauses for minutes | What is actually on the heap, and what is holding it. Last resort. |
Which one answers which question¶
"A channel is running but nothing is moving." View Threads first. A processor thread blocked on a socket read or a database call is visible immediately, and costs nothing to look at. If the picture is unclear or you need to send it to someone, take a Thread Dump.
"Memory keeps climbing." Run GC and watch whether used memory drops. If it does, nothing is wrong. You were looking at garbage that had not been collected yet. If it does not, the memory is genuinely reachable, and a Heap Dump shows what is holding it.
"Qvera support asked for diagnostics." Usually a thread dump, and a heap dump only if they say so. The log export gathers the thread dump files automatically. See Exporting Log Messages.
In a clustered environment¶
Each of these acts on one instance, chosen in the dialog. The request is dispatched to that instance and the resulting file is written on that instance's server. In an EnterpriseHA cluster, that means a dump taken on the wrong node looks perfectly healthy while the problem sits elsewhere.
Identify the affected node first from Global System Status, then target it.
Warning
Thread Dump and Heap Dump both pause message processing on the instance they run against, briefly for a thread dump and for several minutes for a heap dump. Neither is a safe casual click on a production node during business hours. View Threads and Run GC are.