Add migration for groups with invalid values by dhaus67 · Pull Request #3789 · stackrox/stackrox

dhaus67 · 2022-11-11T19:24:34Z

Description

Within the initial PR to migrate the groups bucket from a composite key created by the storage.GroupProperties set on the group to a UUID, groups that had an empty role name have been missed during the migration.

Those have surfaced recently after looking at a ticket (ROX-13446), where the groups bucket contained two groups with properties set to nil and an empty role name.

Irrespective of how the groups got there in the first place, the first migration missed to add an ID to groups with a) either no role name set or b) no properties being set.

After adding validation, groups that either do not reference a role name or an auth provider are considered as invalid.

This PR creates a migration that specifically removes those groups.

Checklist

Investigated and inspected CI test results
Unit test and regression tests added
~~- [ ] Evaluated and added CHANGELOG entry if required~~
~~- [ ] Determined and documented upgrade steps~~
- [ ] Documented user facing changes (create PR based on openshift/openshift-docs and merge into
rhacs-docs)

If any of these don't apply, please comment below.

Testing Performed

see unit tests added for the migration.

Additionally, the following tests were done manually to ensure the migration worked as expected:

Create a central which inserts invalid data on startup (I used a custom build that added those improper groups with nil properties and empty role name).
You should observe the following output when querying the groups endpoint:

roxcurl  /v1/groups
{"groups":[{"props":null,"roleName":""},{"props":{"authProviderId":"177332fb-c6ae-4062-ab60-54b42cdeb024","key":"","value":""},"roleName":""}

Upgrade to this version.
Query the groups endpoint and verify that the previously invalid entries are not available anymore.

roxcurl  /v1/groups
{}

dhaus67 · 2022-11-11T19:24:46Z

Current dependencies on/for this PR:

master
- PR Add migration for groups with invalid values #3789 👈

This comment was auto-generated by Graphite.

ghost · 2022-11-11T19:43:07Z

Images are ready for the commit at 74da504.

To use with deploy scripts, first export MAIN_IMAGE_TAG=3.72.x-621-g74da504ac6.

ivan-degtiarenko · 2022-11-14T16:04:56Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+			// In a prior migration, the groups bucket was migrated from groups being stored by a composite key to groups
+			// being stored by a UUID.
+			// Within that migration, groups that had an empty role name were skipped during migration.
+			// This lead to bucket entries where the properties and role name was both empty, thus making it impossible


Should we also delete groups with empty role names here as a pre-caution?

We do delete groups with empty role names as long as they are still stored keyed by the composite key.

I don't think we need to remove groups with empty role names when they are stored using a unique ID - currently it would be possible to delete those via API directly (I also don't expect those to be there, but then again neither did I with the current situation). Still, I'd say let's not do this and add additional complexity to the migration.

For me, empty role name = invalid group = should be deleted. This corresponds with the goal of this migration as I see it - fix our previous mistake and return the groups bucket to the valid state. Maybe I understood it wrong and we delete somewhere groups with an empty role, but definitely not during startup:

stackrox/central/group/datastore/singleton.go

Lines 22 to 27 in 6113b92

var isEmptyGroupPropertiesF = func(props *storage.GroupProperties) bool {

if props.GetAuthProviderId() == "" && props.GetKey() == "" && props.GetValue() == "" {

return true

}

return false

}

I won't fixate on this though. If you feel strongly about not adding this to migration right now, let's put it on hold - for me, it makes sense to do it now due to our release cadence.

ivan-degtiarenko · 2022-11-14T16:06:44Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+
+			// 1. Remove the value stored behind the composite key, since the migrated group is now successfully stored.
+			if err := bucket.Delete(compositeKey); err != nil {
+				return err


Should we make this migration best-effort by appending an error into multierror instead of returning immediately?

ivan-degtiarenko · 2022-11-14T16:07:25Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+		for i := range groupStoredByCompositeKeys {
+			compositeKey := groupStoredByCompositeKeys[i].compositeKey
+
+			// 1. Remove the value stored behind the composite key, since the migrated group is now successfully stored.


There's no 2

ivan-degtiarenko · 2022-11-14T16:09:42Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+			return nil
+		})
+	})
+	return groupsStoredByCompositeKey, err


let's explicitly return nil instead of err here

ivan-degtiarenko · 2022-11-14T16:12:03Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+			return nil
+		}
+
+		for i := range groupStoredByCompositeKeys {


What if we use for _, group := range groupStoredByCompositeKeys? Then we don't need to address array value by index within the loop

ivan-degtiarenko · 2022-11-14T16:12:46Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration_test.go

+	suite.Run(t, new(removeGroups))
+}
+
+type removeGroups struct {


let's call it removeGroupsMigrationSuite or something similar

ivan-degtiarenko

LGTM modulo couple of nits

ivan-degtiarenko · 2022-11-15T12:42:25Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration_test.go

+		},
+	}
+
+	// 1. Buckets don't exist should succeed still


super-nit

Suggested change

// 1. Buckets don't exist should succeed still

// 1. Migration should succeed if the bucket does not exist

ivan-degtiarenko · 2022-11-15T12:43:16Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/migration.go

+
+		var deleteGroupErrs *multierror.Error
+		for _, group := range groupStoredByCompositeKeys {
+			compositeKey := group.compositeKey


nit: can be inlined

janisz · 2022-11-15T14:51:42Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/binenc/byte_slice_list.go

+}
+
+// EncodeBytesList takes a list of byte slices and encodes them into a single byte slice.
+func EncodeBytesList(byteSlices ...[]byte) []byte {


Who is using it?

I followed the same examples for copying stackrox code into the migrations - which is copying the package as is and not modifying it, even though some functions may be unused.

janisz · 2022-11-15T14:51:50Z

migrator/migrations/m_111_to_m_112_groups_invalid_values/binenc/byte_slice_list.go

+
+// DecodeBytesList takes a byte buffer encoded via EncodeBytesList or WriteBytesList and decodes it into a list of byte
+// slices.
+func DecodeBytesList(buf []byte) ([][]byte, error) {


Who is using it?

Same as above.

7ffc6be ROX-13368: Skip failing nongroovy tests on PG (#3721) bbdd7a0 Bump github.com/gofrs/uuid from 4.3.0+incompatible to 4.3.1+incompatible (#3642) 1f253f2 Bump github.com/google/certificate-transparency-go from 1.1.3 to 1.1.4 (#3543) d434c8d [ROX-13030] : Add delete collection API endpoint and service implementation (#3648) f062c21 Dashrews/ROX-13253 wait for central-db to come back after bounce and allow FATAL connection lost error (#3537) edc1174 CI: Fill the gaps for https://testgrid.k8s.io/ (#3715) 86d7c54 ROX-13231: use passed context when non-postgres (#3540) 9093195 Add less specific type for BE collection response string (#3728) 5abb652 Only enable ROX_OBJECT_COLLECTIONS feature flag during gke-postgres-ui-e2e job (#3727) 4f64cd1 Add centralDBOnly mode in render (#3707) d67bbe5 Dashrews/ROX-13082 UUID searcher and common updates to set allow use of postgres UUID PR 1 of 4 (#3679) 6f829d5 ROX-13259: graphInit called during init time (#3705) f3bc50d ROX-13380: Conditional rendering edges for deployments and namespaces (#3641) 3764476 ROX-12319: implement smoke test step with groovy test filter (#3220) f202fd4 ROX-11826: Disable kernel support package uploads for managed central (#3661) 61f03dc ROX-11101: Remove deprecated resources from central (#3115) e6aa6d7 ROX-11101: Restore Role permission in UI (#3428) 3203e04 ROX-11101: Remove deprecated resources (#3036) a35f41e Bump golang.org/x/sys from 0.1.0 to 0.2.0 (#3733) 4120524 Bump snakeyaml from 1.29 to 1.33 in /qa-tests-backend (#3732) e1785c0 Bump github.com/coreos/go-systemd/v22 from 22.4.0 to 22.5.0 (#3724) 870df4a Bump google.golang.org/api from 0.101.0 to 0.102.0 (#3723) 721454c Generalize User-Agent setup (#3672) 6a11bf0 Bumps collector version to 3.11.x-145-gc345f72f5e (#3736) ec5d343 [ROX-12923] Walk retries - remainder work (#3729) 40f3d43 ROX-13440: Replace ambiguous central with sensor in networkGraph integration test (#3730) 281ed22 Bump groovy-xml from 2.5.18 to 2.5.19 in /qa-tests-backend (#3741) 49d1651 Bump github.com/prometheus/client_golang from 1.13.1 to 1.14.0 (#3742) b5544aa Bump cloud.google.com/go/storage from 1.27.0 to 1.28.0 (#3743) 9c61e53 ensure CVSS is present for istio vulns (#3706) ae29d52 ROX-13452: don't always clobber scoped ctx when non-postgres (#3748) 517bf05 ROX-13261: DryRunUpdate on collection datastore (#3687) baf7654 ROX-13378: Group new resources with deprecated in UI (#3690) 569922f ROX-13421: Enable roxctl netpol generate and add tech-preview messages (#3740) 2465fc5 Dashrews/ROX-13082 UUID generator templates PR 2 of 4 (#3681) c093c68 Bump slack-api-client from 1.20.2 to 1.27.0 in /qa-tests-backend (#3752) 2c860bb Bump ubi8-micro from 8.6 to 8.7 in /operator (#3751) 80eb04c Make deploy.sh and deploy-local.sh pass shellcheck (#3582) 2182b43 Dashrews/ROX-13082 UUID test updates PR 3 of 4 (#3694) 6dc6ca5 [ROX-13403] : Fix node -> topVuln sub resolver bug when node cves is empty (#3689) 1b21361 Move integration tests for page title from general to specific containers (#3675) e1a9f31 Bump google.golang.org/api from 0.102.0 to 0.103.0 (#3773) a05ea31 Bump golang.org/x/crypto from 0.1.0 to 0.2.0 (#3772) 65ddf4f ROX-12824: Add roxctl commands to generate Central DB bundle (#3602) c3f1e2f Remove obsolete authProviders request for Integrations page (#3759) 7ccd54d Dashrews/ROX-13082 UUID protos generated PR 4 of 4 (#3698) 9ab5c8f cleanup image digest utilities (#3764) 187ed44 ROX-11931: Convert junit failure artifacts to Slack attachments (#3438) b5d8790 ROX-13432: leaning up unused code copied/pasted from topology demo (#3750) ab05bfc Refactor collection form page for better composition (#3744) c5562f7 Remove babel devDependencies in ui-components (#3761) 2b90b3a Extract collection form from drawer wrapper layout (#3745) a779fc9 [ROX-12625 + ROX-13032] : Add GetCollectionCount and UpdateCollection endpoints and services (#3749) e77f0da Upgrade cypress 11.0.0 devDependencies in ui (#3760) a3fba94 ROX-13068: Use real data for deployment details (#3688) 4c7d90e ROX-12617: Collection to search query converter (#3683) 3e98aec ROX-13067: fill out port configurations section of deployment details (#3714) a48de36 ROX-12835: Add support for NodeScanV2 to Sensor (#3533) 30c5dc7 ROX-13466: Fix deletion of groups with empty properties (#3756) 5cb2470 Add autocomplete for name selector dropdowns (#3676) b9a75ad ROX-13464 adding flows dropdown in NG (#3763) 3217a67 [ROX-13500] Perform type check for V1 CronJob (#3787) af3790d Remove bulk delete from collections table (#3776) dda123b Add more info in migration log (#3788) 179f0c9 ROX-13502: Remove the circular dependency between cluster datastore init and cscc notifier init (#3790) 029d584 Update SCANNER_VERSION (#3774) cbca57c Bump github.com/ckaznocha/protoc-gen-lint from 0.2.4 to 0.3.0 (#3783) 3613b56 Bump golang.org/x/tools from 0.2.0 to 0.3.0 (#3782) 5fc0a6a Bump github.com/google/go-containerregistry from 0.12.0 to 0.12.1 (#3781) 1d1c687 Bump controller-gen version to 0.10.0 (#3754) c3a5290 Untie documentation link from the product version (#3799) ed822aa use correct package for migration (#3784) 397a0b4 Validate that label keys are valid k8s labels and ensure correct key splitting (#3777) edd1050 Rename variable ScannerGRPCEndpoint to ScannerSlimGRPCEndpoint (#3657) 6662c9f ROX-13378: Access Control page permissions (#3720) b0e73c5 fix Operator reconciliation for external Central DB (#3796) b83bc1f ROX-13505: Fix error log scanning the postgres stat collection (#3795) ca660cb Prevent the collection being edited from displaying in its own embedded list (#3778) 3f7b3fc [ROX-13441][POSTGRES] Propagate context correctly in retries (#3793) e0cbc6f ROX-12839: Update changelog to announce removal of in-product docs (#3805) 696e8bc [ROX-12358] Follow up on vulnerability request proto change (#2851) c4b46d8 Change getCollectionCount endpoint and updateCollection request type 5f2efbc remove make proto-fmt (#3804) 0c75540 Remove os.Std* from roxctl/central (#3758) 25a90de Add ability to view embedded collections in a pop up modal (#3747) 5c1bf81 ROX-13240: fix scanner-slim updates when WebSockets are used (#3704) 1d98577 Add more context to jira notifier logging (#3812) da2fd28 ROX-13031: DryRun Collection API (#3766) 1c418d5 Test data migration code in postgres tests (#3803) ed95b37 Update UI Collection requests for BE compatibiltiy (#3762) 09cc188 ROX-11931: Fix junit-parse install in CI (#3811) d2b01e3 ROX-12814: Disable PolicyFieldsTest on openshift. (#3797) d10ce27 ROX-13345: disable 'missing required registry' aspect on openshift (#3798) 3d22396 Update collector to 3.12 (#3809) 1eb33fb ROX-13347: Modify scope queries to included quoted cluster and nameace names, to allow exact matches instead of erroneous and unintended prefix matches. (#3767) 3811a69 ROX-12621: list collection selectors api (#3806) f6d3f9d Add migration for groups with invalid values (#3789) cc21125 Bugfixes for collection autocomplete (#3816) 7623dec ROX-9350 Use fine-grained host paths for compliance mounts (#2479) b4bf5c2 Fix collector volumeMounts (#3826) 0e9be05 ROX-12953: figure out last 4 versions of sensor automatically (#3611) 459c7ae ROX-12814: Add proper todo for reenabling the test (#3817) 9ee40ff ROX-13523: add isEnabled enum to central db spec (#3815) 535bc72 Replace requestConfig with routeMatcherMap in helper functions for integration tests (#3686) 2b75b61 `gosec` G104: Add `ShouldErr(err)` that returns `err` (#3830) fb1b82f WIP: Introduce nodescan call 35f8a8f WIP: Prepare converter 716144b Moved and renamed fake nodescan tests 4748de7 Introduce real node scanner with conversion functions 5e6d9a8 wip: real scanner 0169868 wip: log results 1438d2c wip: Debug Analyze call b09894c wip: Debug Analyze call 3ceed72 wip: Update and improve debug logs d1669fd Remove copied lib, bump scanner version, add debug 14a3f73 Merge branch 'master' into mm/ROX-12967-real-nodescan fbd0450 Fix style issues 17ccb31 Debug: let both scans finish to see what they return

dhaus67 requested a review from ivan-degtiarenko November 11, 2022 19:24

ivan-degtiarenko reviewed Nov 14, 2022

View reviewed changes

dhaus67 requested a review from ivan-degtiarenko November 15, 2022 01:50

ivan-degtiarenko approved these changes Nov 15, 2022

View reviewed changes

janisz reviewed Nov 15, 2022

View reviewed changes

dhaus67 added 3 commits November 16, 2022 00:56

fix-groups-migration

2a31da4

Address review comments.

a31024b

Address review comments.

74da504

dhaus67 force-pushed the master-dh/fix-groups-migration branch from cc540a9 to 74da504 Compare November 15, 2022 23:56

dhaus67 enabled auto-merge (squash) November 15, 2022 23:57

dhaus67 merged commit f6d3f9d into master Nov 16, 2022

dhaus67 deleted the master-dh/fix-groups-migration branch November 16, 2022 00:30

pedrottimark mentioned this pull request Nov 22, 2022

ROX-13564: Replace vmWorkflowUtils helper functions in non-postgress vulnmanagement integration tests #3885

Merged

2 tasks

dhaus67 mentioned this pull request Dec 8, 2022

ROX-13819: Recreate groups bucket #4068

Merged

5 tasks

	var isEmptyGroupPropertiesF = func(props *storage.GroupProperties) bool {
	if props.GetAuthProviderId() == "" && props.GetKey() == "" && props.GetValue() == "" {
	return true
	}
	return false
	}

	// 1. Buckets don't exist should succeed still
	// 1. Migration should succeed if the bucket does not exist

Conversation

dhaus67 commented Nov 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Checklist

Testing Performed

Uh oh!

dhaus67 commented Nov 11, 2022

Uh oh!

ghost commented Nov 11, 2022 • edited by ghost Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ivan-degtiarenko left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

dhaus67 commented Nov 11, 2022 •

edited

Loading

ghost commented Nov 11, 2022 •

edited by ghost

Loading