From a16207712df8d274f3e4456142d979be3acd7609 Mon Sep 17 00:00:00 2001
From: Jan Teske <jteske@posteo.net>
Date: Wed, 15 Oct 2025 10:58:33 +0200
Subject: [PATCH] testdrive: unflake introspection-sources.td

The queries testing that all workers report the same dataflows,
operators, channels, etc. would get confused by transient dataflows
created for oneshot selects. These dataflows get created and dropped
very quickly, so it is likely for different workers to have different
parts of them installed at any given time. To unflake the queries we
need to filter out transient dataflows.
---
 test/testdrive/introspection-sources.td | 56 ++++++++++++++++++-------
 1 file changed, 40 insertions(+), 16 deletions(-)

diff --git a/test/testdrive/introspection-sources.td b/test/testdrive/introspection-sources.td
index 9bd8b3e038fe6..bb025f86c1acd 100644
--- a/test/testdrive/introspection-sources.td
+++ b/test/testdrive/introspection-sources.td
@@ -156,46 +156,70 @@ $ set-regex match=\d{13} replacement=<TIMESTAMP>
 # Test that certain `*_per_worker` relations include the same data for each
 # worker. The corresponding global relations rely on this to justify showing
 # only the data from worker 0.
+#
+# These queries get confused by transient dataflows that get created and
+# dropped quickly, so we filter these out.
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflows_per_worker
-  GROUP BY id, name
+  FROM mz_introspection.mz_dataflows_per_worker d
+  JOIN mz_introspection.mz_compute_exports e ON d.id = e.dataflow_id
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY d.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflow_addresses_per_worker
-  GROUP BY id, address
+  FROM mz_introspection.mz_dataflow_addresses_per_worker a
+  JOIN mz_introspection.mz_dataflow_operator_dataflows od USING (id)
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY a.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflow_channels_per_worker
-  GROUP BY id, from_index, from_port, to_index, to_port
+  FROM mz_introspection.mz_dataflow_channels_per_worker c
+  JOIN mz_introspection.mz_dataflow_channel_operators co USING (id)
+  JOIN mz_introspection.mz_dataflow_operator_dataflows d ON co.from_operator_id = d.id
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY c.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflow_operators_per_worker
-  GROUP BY id, name
+  FROM mz_introspection.mz_dataflow_operators_per_worker o
+  JOIN mz_introspection.mz_dataflow_operator_dataflows od USING (id)
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY o.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflow_operator_dataflows_per_worker
-  GROUP BY id, name, dataflow_id, dataflow_name
+  FROM mz_introspection.mz_dataflow_operator_dataflows_per_worker od
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY od.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_dataflow_channel_operators_per_worker
-  GROUP BY id, from_operator_id, to_operator_id
+  FROM mz_introspection.mz_dataflow_channel_operators_per_worker co
+  JOIN mz_introspection.mz_dataflow_operator_dataflows d ON co.from_operator_id = d.id
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY co.id
 16
 
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_compute_exports_per_worker
-  GROUP BY export_id, dataflow_id
+  FROM mz_introspection.mz_compute_exports_per_worker e
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY e.export_id
 16
 
 > CREATE DEFAULT INDEX ON t
 > SELECT DISTINCT count(*)
-  FROM mz_introspection.mz_arrangement_sharing_per_worker
-  GROUP BY operator_id, count
+  FROM mz_introspection.mz_arrangement_sharing_per_worker s
+  JOIN mz_introspection.mz_dataflow_operator_dataflows od ON s.operator_id = od.id
+  JOIN mz_introspection.mz_compute_exports e USING (dataflow_id)
+  WHERE e.export_id NOT LIKE 't%'
+  GROUP BY s.operator_id
 16
 
 > DROP INDEX t_primary_idx