* two fixes, compaction and when to run
Test logs are super, duper awesome. We upload them on failure since we
don't presumably need to see healthy logs.
There are two issues:
1. We upload them on `failure()`, but timeouts don't seem to trigger
this condition. So now upload on `always() && !success()` Why not just
`!success()`? An LLM suggests that timeouts are a bit special and go
into another mode. The always() will run and then we qualify it. Let's
see.
2. The logs are all on a single line.
We have a few axes here: compact and eventEol.
I have three files with these. If you want to grep for logs during a
particular test, here's what you get:
Compact: (current behavior)
```
❯ ag "metabase.util.queue-test/result-listener-test" logs/test-log.json.compact
<not reproducing here, because it's the whole file>
```
compact false
```
❯ ag "metabase.util.queue-test/result-listener-test" logs/test-log.json.not-compact
251: "mb-test" : "metabase.util.queue-test/result-listener-test"
268: "mb-test" : "metabase.util.queue-test/result-listener-test"
414: "mb-test" : "metabase.util.queue-test/result-listener-test"
```
It's _too_ expanded.
```
{
"instant" : {
"epochSecond" : 1756844736,
"nanoOfSecond" : 357005000
},
"thread" : "main",
"level" : "INFO",
"loggerName" : "metabase.util",
"message" : "Maximum memory available to JVM: 8.0 GB",
"contextMap" : { },
"endOfBatch" : false,
"loggerFqcn" : "org.apache.logging.log4j.spi.AbstractLogger",
"threadId" : 1,
"threadPriority" : 5
}
{
"instant" : {
"epochSecond" : 1756844739,
"nanoOfSecond" : 802772000
},
```
Compact true, eventEol (this PR sets this)
```
❯ ag "metabase.util.queue-test/result-listener-test" logs/test-log.json
17:{"instant":{"epochSecond":1756844991,"nanoOfSecond":537601000},"thread":"main","level":"INFO","loggerName":"metabase.util.queue","message":"Starting listener \u001B[32mtest-result-listener\u001B[0m with 1 threads 🎧","contextMap":{"mb-test":"metabase.util.queue-test/result-listener-test"},"endOfBatch":false,"loggerFqcn":"org.apache.logging.log4j.spi.AbstractLogger","threadId":1,"threadPriority":5}
18:{"instant":{"epochSecond":1756844991,"nanoOfSecond":572361000},"thread":"main","level":"INFO","loggerName":"metabase.util.queue","message":"Stopping listener test-result-listener...","contextMap":{"mb-test":"metabase.util.queue-test/result-listener-test"},"endOfBatch":false,"loggerFqcn":"org.apache.logging.log4j.spi.AbstractLogger","threadId":1,"threadPriority":5}
20:{"instant":{"epochSecond":1756844991,"nanoOfSecond":585407000},"thread":"main","level":"INFO","loggerName":"metabase.util.queue","message":"Stopping listener test-result-listener...done","contextMap":{"mb-test":"metabase.util.queue-test/result-listener-test"},"endOfBatch":false,"loggerFqcn":"org.apache.logging.log4j.spi.AbstractLogger","threadId":1,"threadPriority":5}
```
You get the whole log lines. super helpful. pipe these through a
processing pipeline, use jq, etc.
Also, these files are ~817MB when running in a test suite. I'm doing these with a single namespace
```
❯ clj -X:dev:test :only metabase.util.queue-test
❯ ls -lh logs
13K Sep 2 15:29 test-log.json
13K Sep 2 15:24 test-log.json.compact
16K Sep 2 15:26 test-log.json.not-compact
```
* stacktraces when we don't get returned connections
9f6acff3f5