OpenSearch upgrade leftover persistent settings

Plugins Index Management
1

Hello,

After OpenSearch upgrade we note that these entries have been added to the persistent settings of the clusters:

{
  "persistent" : {
    "plugins" : {
      "index_state_management" : {
        "metadata_migration" : {
          "status" : "1"
        },
        "template_migration" : {
          "control" : "-1"
        }
      }
    }
...
}

I haven’t found any documentation around them but I guess they were used to coordinate the migration. May I delete them now that clusters are happily upgraded? Is there any documentation around these that I missed? Many thanks in advance.

2

Have the same question, anybody has more information about this?

3

following, any updates?

4

Hi,
I join the question

In a freshly deployed cluster, there is only one option:

GET _cluster/settings?flat_settings
...
{
   "persistent": {
     "plugins.index_state_management.template_migration.control": "-1"
   },
   "transient": {}
}

In the cluster log I constantly see the lines

[2023-03-07T06:31:48,899][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node-data201] Detected cluster change event for destination migration
[2023-03-07T06:35:03,630][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node-data201] Detected cluster change event for destination migration
[2023-03-07T06:35:04,182][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node-data201] Detected cluster change event for destination migration

In existing clusters, two parameters as @spapadop wrote and the same lines in the log, although all nodes have long been updated and work

5

Yes, I also tried to remove that setting with:

PUT _cluster/settings
{
  "persistent": {
    "plugins": {
      "index_state_management": {
        "metadata_migration": {
          "status": null
        },
        "template_migration": {
          "control": null
        }
      }
    }
  }
}

And indeed, after a while plugins.index_state_management.template_migration.control comes back:

{
  "persistent": {
    "plugins": {
      "index_state_management": {
        "metadata_migration": {
          "status": "1"
        },
        "template_migration": {
          "control": "-1"
        }
      }
    }
  },
  "transient": {}
}

And as @mouse said, these destination migration logs never disappear:

[2023-09-01T14:58:41,560][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
[2023-09-01T14:58:42,054][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
[2023-09-01T14:58:42,072][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
[2023-09-01T14:59:38,094][INFO ][o.o.c.s.ClusterSettings  ] [node1] updating [plugins.index_state_management.template_migration.control] from [0] to [-1]
[2023-09-01T14:59:38,097][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
[2023-09-01T15:01:08,007][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
[2023-09-01T15:01:51,324][INFO ][o.o.c.s.ClusterSettings  ] [node1] updating [plugins.index_state_management.template_migration.control] from [-1] to [0]
[2023-09-01T15:01:51,327][INFO ][o.o.a.u.d.DestinationMigrationCoordinator] [node1] Detected cluster change event for destination migration
6

I have opened a github issue so that this is followed up there.