Tablet move and group removal

[srfrog]

This PR adds tests to the process of moving a tablet to another group, and removing nodes from a group. We use this to test that these processes don't break the continuity of the cluster.

Test methodology:

1. Load test data into 3-group cluster
2. Move all predicates from group 3 to group 2
3. Remove group 3
4. Run a test query using the data.
5. Move all predicates from group 2 to group 1
6. Remove group 2
7. Run a test query using the data, must match step 4.

Closes #2466

---

This change is 

[codexnull]

Reviewable status: 0 of 4 files reviewed, 1 unresolved discussion (waiting on @srfrog, @gitlw, @manishrjain, and @codexnull)

---

systest/nodes_test.go, line 97 at r1 (raw file):

			})
			x.Check(err)
			cnt = 0

cnt = 0 is not necessary since you're using cnt%100 above.

[srfrog]

Reviewable status: 0 of 4 files reviewed, 1 unresolved discussion (waiting on @codexnull, @gitlw, and @manishrjain)

---

systest/nodes_test.go, line 97 at r1 (raw file):

Previously, codexnull (Javier Alvarado) wrote…

cnt = 0 is not necessary since you're using cnt%100 above.

nice catch, thx!

[manishrjain]

Got a comment. So, address that before merging.

Reviewed 3 of 4 files at r1, 1 of 1 files at r2.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @codexnull, @srfrog, and @gitlw)

---

systest/nodes_test.go, line 141 at r2 (raw file):

	for pred := range state1.Groups["3"].Tablets {
		url := fmt.Sprintf("http://localhost:6080/moveTablet?tablet=%s&group=2", pred)
		resp, err := http.Get(url)

resp could also have some errors in there. Typically, that's what we do. So, check for those.

[srfrog]

Reviewable status: 3 of 4 files reviewed, 2 unresolved discussions (waiting on @manishrjain, @codexnull, and @gitlw)

---

systest/nodes_test.go, line 141 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

resp could also have some errors in there. Typically, that's what we do. So, check for those.

Done.

[srfrog]

Reviewable status: 3 of 4 files reviewed, 2 unresolved discussions (waiting on @manishrjain, @codexnull, and @gitlw)

---

systest/nodes_test.go, line 97 at r1 (raw file):

Previously, srfrog (Gus) wrote…

nice catch, thx!

Done.

Outdated

[kuberxy]

This PR adds tests to the process of moving a tablet to another group, and removing nodes from a group. We use this to test that these processes don't break the continuity of the cluster.

Test methodology:

1. Load test data into 3-group cluster
2. Move all predicates from group 3 to group 2
3. Remove group 3
4. Run a test query using the data.
5. Move all predicates from group 2 to group 1
6. Remove group 2
7. Run a test query using the data, must match step 4.

Closes #2466

This change is 

Hello, how exactly do you do when you remove a group?

There are 4 groups in my cluster.

After I removed all the instances in group 4(by curl "http://127.0.0.1:6080/removeNode?group=4&id=10" ), my cluster became read-only.

When I write data, I get an error like this

{"errors":[{"message":"mutation addProduct failed because Dgraph execution failed because cannot retrieve predicate information: No connection exists","locations":[{"line":2,"column":3}],"path":["addProduct"]},{"message":"Mutation addCustomer was not executed because of a previous error.","locations":[{"line":11,"column":3}],"path":["addCustomer"]}],"data":{"addProduct":null},"extensions":{"tracing":{"version":1,"startTime":"2024-09-05T14:44:38.672030942+08:00","endTime":"2024-09-05T14:44:38.675771422+08:00","duration":3740518,"execution":{"resolvers":[{"path":["addProduct"],"parentType":"Mutation","fieldName":"addProduct","returnType":"AddProductPayload","startOffset":209005,"duration":3487989,"dgraph":[{"label":"preMutationQuery","startOffset":0,"duration":0},{"label":"mutation","startOffset":277442,"duration":0},{"label":"query","startOffset":0,"duration":0}]}]}}}}

My question is similar to this https://discuss.dgraph.io/t/option-to-gracefully-remove-a-undead-group/2921/5