Skip to content

Replace MinIO with SeaweedFS in xbstream_fifo_test.sh, fix keyring_file bugs - #11

Merged
saikumar-vs merged 1 commit into
mainfrom
pxb-seaweedfs-migration
Sep 22, 2026
Merged

Replace MinIO with SeaweedFS in xbstream_fifo_test.sh, fix keyring_file bugs#11
saikumar-vs merged 1 commit into
mainfrom
pxb-seaweedfs-migration

Conversation

@saikumar-vs

@saikumar-vs saikumar-vs commented Sep 22, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Summary

  • Swap start_minio() for start_seaweedfs() in xbstream_fifo_test.sh: runs SeaweedFS's S3 gateway via docker (chrislusf/seaweedfs), publishing container port 8333 as host 9000 so every existing xbcloud --s3-endpoint=http://localhost:9000 call is unchanged. Generates an S3 identity config so the admin/password credentials xbcloud already uses keep working. Readiness check polls the S3 port directly since SeaweedFS has no MinIO-style /minio/health/ready endpoint. cleanup_exit() now stops the seaweedfs container on script exit (success, error, or interrupt).
  • Restore test_scripts/pxb/kmip_helper.sh (it existed under test_scripts/pxb/old/ but was never re-pathed back after a prior file reorg), which init_datadir() sources for the keyring_kmip scenarios.
  • Fix full_backup_and_restore(): $COMPRESS_OPTIONS was set by the compressed-backup scenario but never actually passed to xtrabackup --backup, so compression never happened.
  • Fix init_datadir(): component_keyring_file.cnf used the wrong JSON key (component_keyring_file_data instead of path), so the keyring component silently failed to initialize.
  • Fix incremental_backup_and_restore(): the xtrabackup --backup calls never passed keyring config, so encrypted backups failed to init the keyring component during backup (--prepare already passed it correctly).

Test plan

  • Full backup/restore over SeaweedFS (Test 1) — verified against real PXB 8.4.0-7 / PS 8.4.10-10 builds
  • Incremental backup (full + 3 incrementals) over SeaweedFS (Test 2) — verified
  • Compressed backup over SeaweedFS, with the $COMPRESS_OPTIONS fix (Test 3) — verified, confirmed real zstd compression in logs
  • Partition-table backup via pstress over SeaweedFS (Test 4) — verified in isolation (the first attempt, chained after Tests 1-3, hit a FIFO race under very high cumulative file count; did not reproduce on a fresh, isolated run)
  • cleanup_exit() stops the SeaweedFS container on exit — verified directly

@saikumar-vs @claude
Replace MinIO with SeaweedFS in xbstream_fifo_test.sh, fix keyring_fi…
36d19be 
…le bugs

- Swap start_minio() for start_seaweedfs(): runs SeaweedFS's S3 gateway via
  docker (chrislusf/seaweedfs), publishing container port 8333 as host 9000
  so all existing xbcloud --s3-endpoint=http://localhost:9000 calls are
  unchanged. Generates an S3 identity config so the admin/password creds
  xbcloud already uses keep working. Readiness check polls the S3 port
  directly since SeaweedFS has no MinIO-style /minio/health/ready endpoint.
  cleanup_exit() now stops the seaweedfs container on script exit.
- Restore test_scripts/pxb/kmip_helper.sh (present in test_scripts/pxb/old/
  but never re-pathed after a prior reorg), which init_datadir() sources
  for the keyring_kmip scenarios.
- Fix full_backup_and_restore(): $COMPRESS_OPTIONS was set by the
  compressed-backup scenario but never passed to xtrabackup --backup, so
  compression never actually happened.
- Fix init_datadir(): component_keyring_file.cnf used the wrong JSON key
  (component_keyring_file_data instead of path), so the keyring component
  silently failed to initialize.
- Fix incremental_backup_and_restore(): the xtrabackup --backup calls never
  passed keyring config, so encrypted backups failed to init the keyring
  component during backup (prepare already passed it correctly).

Verified full/incremental/compressed/partition-table backups against
SeaweedFS end to end. Encrypted (keyring_file) backup now gets past both
keyring bugs but hits an unrelated InnoDB assertion crash inside
xtrabackup itself (fil0fil.cc:10943) during incremental backup of an
encrypted tablespace - a separate PXB issue to track, not a script bug.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@saikumar-vs saikumar-vs mentioned this pull request Sep 22, 2026
Draft
9 tasks
@saikumar-vs
saikumar-vs merged commit f4fe085 into main Sep 22, 2026
3 checks passed
saikumar-vs added a commit that referenced this pull request Sep 22, 2026
@saikumar-vs @claude
Enable hashicorp KMIP vault type, fix broken script interpreter
96c8151 
hashicorp was commented out of KMIP_CONFIGS/DEFAULT_KMIP_CONFIGS and, even
if enabled, KMIPHelper.setup_hashicorp() executed the downloaded
hashicorp_kmip_setup.py (a real Python script, #!/usr/bin/env python3,
argparse-based) via `bash -s`, which would have failed immediately with a
syntax error. Fixed to invoke it via `python3 -`, matching how the retired
bash kmip_helper.sh ran it.

Enabling hashicorp in the shared KMIP_CONFIGS adds a `[hashicorp]`
parametrized variant to every vault_type-parametrized test that reads from
it (test_fifo_kmip_backup here, plus the existing test_kmip_component_backup
and test_crash_backup_encrypted_kmip in inc_backup_load_tests.py). Added a
VAULT_LICENSE environment-variable guard (pytest.skip when unset) at each
of those three call sites, matching the existing FORTANIX_EMAIL/PASSWORD
guard pattern, since hashicorp_kmip_setup.py requires a real HashiCorp
Vault Enterprise license.

Verified: xbstream_fifo_tests.py::test_fifo_kmip_backup[hashicorp] skips
cleanly (no VAULT_LICENSE in this environment) instead of failing; full
license-driven setup path still needs a real Vault Enterprise license to
verify end-to-end, same limitation noted for the bash script in PR #11.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
saikumar-vs added a commit that referenced this pull request Sep 22, 2026
@saikumar-vs @claude
Port xbstream_fifo_test.sh to a pytest module (xbstream_fifo_tests.py)
a7dd75c 
Adds FIFO-streamed cloud backup coverage to the Python test framework,
following the conventions of innodb_myrocks_backup_tests.py and
inc_backup_load_tests.py:

- test_helper.py: additive FIFO_STREAM/FIFO_DIR constants and six new
  BackupTestHelper methods (take_fifo_backup_step, restore_fifo_backup_step,
  take_fifo_full_backup_and_restore, take_fifo_incremental_backup_and_restore,
  restore_datadir_from, cleanup_fifo_state). These stream xtrabackup/xbstream
  and xbcloud through named pipes as two concurrent processes (background +
  foreground), mirroring the bash script's xbcloud_put()/xbcloud_get(),
  rather than take_backup()'s "cloud" backup_type (a single shell pipe).
  No existing method was modified.
- seaweedfs_helper.py: new SeaweedFSHelper class managing the local
  SeaweedFS S3-gateway docker container, ported from the already-fixed
  xbstream_fifo_test.sh start_seaweedfs()/cleanup_exit() shell logic
  (percona-qa/server-qa PR #11).
- xbstream_fifo_tests.py: the actual test file, with a session-scoped
  seaweedfs fixture and 7 tests mirroring the bash script's scenarios
  (full, incremental, compressed, partition-table, keyring_file, keyring_kmip,
  xbcrypt-encrypted backup).
- README.md: new section documenting env vars, how to run, and the test
  reference table, matching the existing per-file structure.

Verified end-to-end against real PXB 8.4.0-7 / PS 8.4.10-10 builds and a
live SeaweedFS container: all 7 tests pass (test_fifo_kmip_backup requires
KMIP vault/license resources not available in this environment, so it
wasn't run, matching the same gap noted in the bash-script PR). One
transient SeaweedFS-readiness timeout and one missing --xtrabackup-plugin-dir
in two tests' prepare_params were found and fixed during that verification.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
saikumar-vs added a commit that referenced this pull request Sep 22, 2026
@saikumar-vs @claude
Enable hashicorp KMIP vault type, fix broken script interpreter
f15e652 
hashicorp was commented out of KMIP_CONFIGS/DEFAULT_KMIP_CONFIGS and, even
if enabled, KMIPHelper.setup_hashicorp() executed the downloaded
hashicorp_kmip_setup.py (a real Python script, #!/usr/bin/env python3,
argparse-based) via `bash -s`, which would have failed immediately with a
syntax error. Fixed to invoke it via `python3 -`, matching how the retired
bash kmip_helper.sh ran it.

Enabling hashicorp in the shared KMIP_CONFIGS adds a `[hashicorp]`
parametrized variant to every vault_type-parametrized test that reads from
it (test_fifo_kmip_backup here, plus the existing test_kmip_component_backup
and test_crash_backup_encrypted_kmip in inc_backup_load_tests.py). Added a
VAULT_LICENSE environment-variable guard (pytest.skip when unset) at each
of those three call sites, matching the existing FORTANIX_EMAIL/PASSWORD
guard pattern, since hashicorp_kmip_setup.py requires a real HashiCorp
Vault Enterprise license.

Verified: xbstream_fifo_tests.py::test_fifo_kmip_backup[hashicorp] skips
cleanly (no VAULT_LICENSE in this environment) instead of failing; full
license-driven setup path still needs a real Vault Enterprise license to
verify end-to-end, same limitation noted for the bash script in PR #11.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@tplavcic tplavcic tplavcic approved these changes
@hrvojem hrvojem Awaiting requested review from hrvojem
@mchawla16 mchawla16 Awaiting requested review from mchawla16

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@saikumar-vs @tplavcic