Add new vshard configuration options

[yngvar-antonsson]

I didn't forget about

• Tests
• Changelog
• Documentation

Close #2185

[RunsFor]

Thanks for a quick PR!

I'm wondering, did you test rebalancer option with the vshard version >= 0.1.25?

I didn't see auto tests in the PR, and also it seems, you've added rebalancer option on top of vshard config, when it should be on either replicaset or replica level.

Here is the link to the commit which adds support for rebalancer option.

[RunsFor]

Hi! I cannot figure out how can I simply disable rebalancer with rebalancer_mode = 'off' option. It would be great to be able to do it in cartridge.cfg({rebalancer_mode = 'off'})

[RunsFor]

What about adding ability to manipulate this option before bootstrapping? the same as with bucket_count. In my case, I would like to bootstrap cluster knowing rebalancer is disabled

cartridge/cartridge.lua

Lines 782 to 793 in bcb7eae

	-- Set up vshard groups
	if next(vshard_groups) == nil then
	vshard_groups = nil
	else
	for _, params in pairs(vshard_groups) do
	if params.bucket_count == nil then
	params.bucket_count = opts.bucket_count
	end
	end
	end
	vshard_utils.set_known_groups(vshard_groups, opts.bucket_count)

[RunsFor]

Thanks!

I've checked rebalancer_mode option. It seems to work well.

But I had some issues with rebalancer option. I've tried to set rebalancer option to true to the working cluster using this query:

{
	"query": "mutation editTopology($replicasets:[EditReplicasetInput]){cluster{edit_topology(replicasets:$replicasets){replicasets{uuid weight roles alias rebalancer}}}}",
	"variables": {
		"replicasets": [
			{
				"weight": 0,
				"uuid": "ebef4119-7e39-41d5-a684-6f99e83c98b0",
				"alias": "core-1",
				"rebalancer": true
			}
		]
	}
}

The query was successful, but the response was wrong:

{
    "data": {
        "cluster": {
            "edit_topology": {
                "replicasets": [
                    {
                        "alias": "core-1",
                        "rebalancer": null,
                        "roles": [
                            "vshard-router",
                            "vshard-storage",
                            "app.roles.sharding-storage",
                            "app.roles.queue"
                        ],
                        "uuid": "ebef4119-7e39-41d5-a684-6f99e83c98b0",
                        "weight": 0
                    }
                ]
            }
        }
    }
}

Check out the "rebalancer": null part

It seems that it has not been passed to graphql Replicaset object since it is actually has been saved into cartridge cw config:

r-1> cartridge.config_get_deepcopy('topology').replicasets['ebef4119-7e39-41d5-a684-6f99e83c98b0']
---
- weight: 0
  vshard_group: default
  alias: core-1
  master:
  - 20f05d26-0560-4b8b-848f-90558f7e496d
  all_rw: false
  roles:
    app.roles.sharding-storage: true
    app.roles.queue: true
    vshard-storage: true
    ddl-manager: true
    metrics: true
    vshard-router: true
  rebalancer: true
...

I suppose you need to add rebalancer field to the replicaset in get-topology module, somewhere here:

cartridge/cartridge/lua-api/get-topology.lua

Lines 104 to 125 in d1a11f4

	replicasets[replicaset_uuid] = {
	uuid = replicaset_uuid,
	roles = {},
	status = 'healthy',
	master = {
	uri = 'void',
	uuid = 'void',
	status = 'void',
	message = 'void',
	},
	active_master = {
	uri = 'void',
	uuid = 'void',
	status = 'void',
	message = 'void',
	},
	weight = nil,
	vshard_group = replicaset.vshard_group,
	servers = {},
	all_rw = replicaset.all_rw or false,
	alias = replicaset.alias or 'unnamed',
	}

And here is get_replicasets call lacking rebalancer field:

r-1> lua_api_topology.get_replicasets('ebef4119-7e39-41d5-a684-6f99e83c98b0')
---
- - &0
    active_master: &1
      disabled: false
      message:
      labels: []
      electable: true
      uuid: 20f05d26-0560-4b8b-848f-90558f7e496d
      uri: localhost:3303
      alias: s-1
      clock_delta: 2.3e-05
      replicaset: *0
      priority: 1
      status: healthy
    master: *1
    status: healthy
    all_rw: false
    vshard_group: default
    alias: core-1
    weight: 0
    servers:
    - *1
    roles:
    - vshard-router
    - vshard-storage
    - app.roles.sharding-storage
    - app.roles.queue
    uuid: ebef4119-7e39-41d5-a684-6f99e83c98b0
...

[RunsFor]

LGTM, thanks!