This is an automated email from the ASF dual-hosted git repository.
gfphoenix78 pushed a commit to branch sync-with-upstream
in repository https://gitbox.apache.org/repos/asf/cloudberry-gpbackup.git
The following commit(s) were added to refs/heads/sync-with-upstream by this
push:
new af1fedf7 fix(test): Skip resize cluster test for Cloudberry due to
replicated table issue (#36)
af1fedf7 is described below
commit af1fedf7c0ddda9ddea7ec7114210f9d90593302
Author: Robert Mu <[email protected]>
AuthorDate: Fri Sep 5 15:17:17 2025 +0800
fix(test): Skip resize cluster test for Cloudberry due to replicated table
issue (#36)
Skip resize cluster test in end-to-end suite for Cloudberry due to
COPY FROM ON SEGMENT issue with replicated tables. Add detailed TODO
comment in restore/data.go explaining the root cause of row count
validation failures during resize operations.
References Cloudberry issue #1298 for tracking the upstream fix.
Ensures test stability while maintaining compatibility with
Greenplum behavior.
The test is temporarily disabled for Cloudberry until the upstream
issue with COPY command returning incorrect row counts for replicated
tables is resolved.
See: https://github.com/apache/cloudberry/issues/1298
---
end_to_end/end_to_end_suite_test.go | 8 ++++++++
restore/data.go | 7 +++++++
2 files changed, 15 insertions(+)
diff --git a/end_to_end/end_to_end_suite_test.go
b/end_to_end/end_to_end_suite_test.go
index af9b575c..2e9ef036 100644
--- a/end_to_end/end_to_end_suite_test.go
+++ b/end_to_end/end_to_end_suite_test.go
@@ -2288,6 +2288,14 @@ LANGUAGE plpgsql NO SQL;`)
if useOldBackupVersion {
Skip("Resize-cluster was only
added in version 1.26")
}
+
+ // TODO: Re-enable this test for
Cloudberry once the following issue is fixed:
+ //
https://github.com/apache/cloudberry/issues/1298
+ if restoreConn.Version.IsCBDB() {
+ Skip(`This test is skipped for
Cloudberry due to an issue with COPY FROM ON SEGMENT for
+ replicated tables
(https://github.com/apache/cloudberry/issues/1298).`)
+ }
+
extractDirectory :=
extractSavedTarFile(backupDir, tarBaseName)
defer
testhelper.AssertQueryRuns(restoreConn, `DROP SCHEMA IF EXISTS schemaone
CASCADE;`)
diff --git a/restore/data.go b/restore/data.go
index 18c0bcd1..2709af18 100644
--- a/restore/data.go
+++ b/restore/data.go
@@ -140,6 +140,13 @@ func restoreSingleTableData(fpInfo *filepath.FilePathInfo,
entry toc.Coordinator
numRowsRestored /= int64(destSize)
}
+ // TODO: When restoring a replicated table to a cluster with a
different number of segments (resize),
+ // this check may fail for Cloudberry. This is because gprestore
assumes that the COPY
+ // command for a replicated table returns the total number of rows
copied across all segments
+ // (N * rows_per_segment) and divides the result by N to get the actual
row count.
+ // However, due to Cloudberry issue
https://github.com/apache/cloudberry/issues/1298,
+ // its COPY ON SEGMENT command returns only the base row count
(rows_per_segment).
+ // The subsequent division in gprestore leads to a miscalculated row
count, causing this check to fail.
err := CheckRowsRestored(numRowsRestored, numRowsBackedUp, tableName)
if err != nil {
gplog.Error(err.Error())
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
