Similar to `tenant configure`, this PR adds `ignore_capacity_limit` as an optional argument to `tenant create`.
This allows the user of fdbcli to create a new tenant on an **assigned** cluster, ignoring the tenant group capacity
on that specific cluster.
When creating a tenant with `ignore_capacity_limit`.
- If the user does not specify `assigned_cluster`, this is an error.
- If the user specifies `assigned_cluster`,
- user does not specify `tenant_group`, then the new tenant will be an ungrouped tenant on the `assigned_cluster` ignoring the capacity limit
- user specifies `tenant_group`,
- if `tenant_group` does not exist, then the new tenant will be created on the assigned cluster and the tenant group will be implicitly created.
- if `tenant_group` already exists, then additional check will make sure the tenant_group's cluster matches what the user specifies.
Test plan:
Simulation and metacluster_fdbcli_tests.py
---------
Co-authored-by: A.J. Beamon <aj.beamon@snowflake.com>
In MetaclusterManagementWorkload, if we lock a tenant that has already been locked with a different lockId, it's possible that a tenant_locked error is thrown. We need to account for this instead of terminating the test.
Test plan:
Regular simulation and a deterministic reproduction
Branch: https://github.com/sfc-gh-yajin/foundationdb/tree/snow-791039
Commit hash: 7356504723
Profile: team
Command: devRetryCorrectnessTest bin/fdbserver -r simulation -f tests/slow/MetaclusterManagement.toml -s 3475230174 -b off --crash --trace_format json
This PR adds auto_tenant_assignment option to register/configure data clusters.
Setting auto_tenant_assignment to disabled means the data cluster is a dedicated one and won't be
used for auto tenant assignment. This option is enabled by default (allowing auto tenant assignment).
Test plan:
simulation tests and metacluster_fdbcli_tests.py
---------
Co-authored-by: A.J. Beamon <aj.beamon@snowflake.com>
1. When retrying the transaction to register a restoring cluster, don't choose a new ID if the current ID matches the one recorded for the restoring cluster
2. A metacluster test was incorrectly handling the case where a transaction was retried with unknown result and had committed successfully
* initial commit to split tenant group metadata
* attempt to fix merge errors
* fix compile errors and adjust existing tests
* fix infinite loop and extra ACTOR tag
* direct assignment instead of store
* direct assign instead of store (missed a few)
Since we currently do not support tenant movement, we should as well explicitly disallow changing the assigned
cluster of a tenant during configuration by throwing `invalid_tenant_configuration` for now.
Test plan:
- add coverage for changing assigned cluster during tenant configuration
- fdbcli
- simulation tests
A.J. Beamon