roachtest: backport improvements to mixedversion and backup test
#103912
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit updates the signature of functions in the `mixedversion` package that create steps that run in the background; this includes functions in the `Helper` struct, as well as in the main `mixedversion.Test` struct. These functions now all return a `mixedversion.StopFunc` that test authors can call when they wish to stop a background function without causing the test to fail. When the stop function is called, the context passed to the background functions is canceled; however, that context cancelation is captured by the framework and logged as an expected termination. If the context is canceled through other means, the test will fail as usual. Epic: CRDB-19321 Release note: None
This adds support to database and full cluster backups in the `backup/mixed-version` roachtest. When a cluster is taken in mixed-version state, a randomly chosen backup type is performed. Since cluster backups also backup some system tables, this commit also adds logic related to verifying that the contents of system tables are properly preserved when these cluster backups are restored. Epic: CRDB-19321 Release note: None
This commit adds a second workload to the `backup/mixed-versions` test, namely `tpcc` with 100 warehouses. Previously, we would only run the `bank` workload; by running TPCC, we verify backups of databases that includes multiple tables, and the backup of tables that includes foreign key constraints. It should also put the cluster under more load than just `bank`, potentially exposing interesting edge cases that wouldn't come up before. Epic: CRDB-19321 Release note: None
This commit updates the `backup/mixed-version` (now `backup-restore/mixed-version`) to also perform mixed-version restores. At a high level, it introduces a new function that runs in mixed-version state that will randomly attempt to restore a subset of the backups taken up to that point during the test. The test will verify that we are able to restore, in mixed version, backups taken in the previous version _and_ in mixed version. Resolves: cockroachdb#96367. Epic: CRDB-19321 Release note: None
This commit makes some final (for now) changes to the `backup-restore/mixed-version` roachtest. Specifically: * we set some backup/restore related cluster settings. These are publicly documented settings and should help expose corner cases that might be harder to come up naturally using the default settings. This is an area that is known to need more tests, as described in a recent postmortem [1]. * introduce a background function that executes statements that lead to rows being inserted into system tables, particularly those that are generally empty in most tests. * simplify the workload setup in the test: the `bank` workload is responsible for testing edge cases, while `tpcc` is a workload that should better represent customer workloads. * verify that backups taken in mixed-version can be restored both in the previous version and in the next version. Previously, we were only testing the next version. Note that most of these changes are not specificaly related to the mixed-version context this test is in. In the future, these features should be packaged in a format that is easier to consume by other tests. [1] https://cockroachlabs.atlassian.net/wiki/spaces/ENG/pages/3013804060/Postmortem+101963+revision+history+backups Epic: none Release note: None
When a mixed-version test (using the `mixedversion` package) starts, the `cockroach` binary is uploaded to every node, and then the cluster is started. If, for some reason, the cockroach process crashes in this startup phase, the `mixedversion` package would panic while generating an error message. The reason for that is that there was an assumption that the connection cache was initialized at that point, which does not hold when the test failure happened on test setup. This fixes this issue by making sure we check for the status of the connection cache when generating error messages. We also make sure concurrent accesses to the connection cache are safe; while this is not strictly necessary (no concurrent reads and writes to it right now), it will likely help in the future as this code changes. Epic: none Release note: None
renatolabs
requested review from
herkolategan and
smg260
and removed request for
a team
|
This change is |
herkolategan
approved these changes
May 25, 2023
srosenberg
approved these changes
May 26, 2023
|
Ran the The failure is a documented bug -- #103597. Merging, TFTR! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Backports PRs:
mixedversionwhen cockroach fails to start" (roachtest: fix panic inmixedversionwhen cockroach fails to start #103799)Please see individual PRs for details.