-
Task
-
Resolution: Done
-
Major
-
None
-
None
-
5
-
False
-
-
True
-
-
The solution for SWATCH-2712 resolved the immediate bug with an exception being thrown, but the solution results in there being leftover buckets and Cores/Sockets measurements in the database during a specific use case. This can cause duplicate hosts to be returned via the Instance API, resulting in a duplicates shown in the GUI for an affected product.
Nightly tally creates an rhsm host record with HBI_HOST as instance_type. It has corresponding host_tally_buckets and Cores/Sockets measurements.
| Host Records | ||
|---|---|---|
| id | instance_id | instance_type |
| host_id_1 | instance_abc | HBI_HOST |
| instance_measurements | ||
| host_id | metric_id | value |
| host_id_1 | Cores | 1 |
| host_id_1 | Sockets | 1 |
Event records gets created in swatch that has a timestamp after the host.last_seen timestamp. Hourly tally happens for that same instance. The host record is going to get UPDATED, and the instance_type will get set to RHEL System. Additional host_tally_buckets and instance_measurements will be created for the relevant metric_id the hourly tally is for.
You'll have one RHEL System host record, and corresponding host_tally_buckets for Cores, Sockets, new metric_Ids.
| Host Records | ||
|---|---|---|
| id | instance_id | instance_type |
| host_id_1 | instance_abc | RHEL System |
| instance_measurements | ||
| host_id | metric_id | value |
| host_id_1 | Cores | 1 |
| host_id_1 | Sockets | 1 |
| host_id_1 | vCPUs | 1 |
The next nightly tally will see that there's no host for that instance with HBI_HOST instance_type, and will create an additional host record for the instance_id. That host record will get new associated host_tally_buckets and instance_measurements for Cores and Sockets again.
The database will the contain
| Host Records | ||
|---|---|---|
| id | instance_id | instance_type |
| host_id_1 | instance_abc | RHEL System |
| host_id_2 | instance_abc | HBI_HOST |
| instance_measurements | ||
| host_id | metric_id | value |
| host_id_1 | Cores | 1 |
| host_id_1 | Sockets | 1 |
| host_id_1 | vCPUs | 1 |
| host_id_2 | Cores | 1 |
| host_id_2 | Sockets | 1 |
The Cores/Sockets measurements should not be tracked with the host_id_1 id anymore, since its host record is no longer HBI_HOST.
This liquibase should rectify those discrepancies, and result in end state of the database being something like this:
| Host Records | ||
|---|---|---|
| id | instance_id | instance_type |
| host_id_1 | instance_abc | RHEL System |
| host_id_2 | instance_abc | HBI_HOST |
| instance_measurements | ||
| host_id | metric_id | value |
| host_id_1 | vCPUs | 1 |
| host_id_2 | Cores | 1 |
| host_id_2 | Sockets | 1 |
