…6239)

* Fix URL paths (#6189)

(cherry picked from commit 4ee4af8)

* fix deploy page to not show blank page

(cherry picked from commit 89e012b)

Co-authored-by: Apoorv Vardhan <vardhanapoorv@users.noreply.github.com>

* Port GraphQL docs to v20.07

This reverts commit 8db61a3.

Co-authored-by: Pawan Rawal <pawan0201@gmail.com>

Passing a few variables to the hugo theme

Co-authored-by: Sankalan Parajuli <sankalan.13@gmail.com>

…6259)

This change also adds a way to set the table and value log loading modes for the w directory
independently of the values for the p directory.

Fixes DGRAPH-1898.

(cherry picked from commit 5f5aa9c)

…6272) (#6276)

Currently a node is marked as visited if it's been visited before AND
has been visited from the same source UID. In dense graphs, the second
condition leads to exponential growth of the data and to other issues
such as responses that are too big to encode.  Removing this condition
fixes the issue.

Fixed tests and verified the new output makes sense.

Fixes DGRAPH-2337

(cherry picked from commit 3638c12)

(cherry picked from commit 52e136f)

)

This was being done in the S3 handler but not in the file handler.
By default the paths are sorted by the filename but sorting by backup
number is more robust.

(cherry picked from commit df63d6c)

) (#6298)

(cherry picked from commit 6f883a4)

…eries. (#6272) (#6276)" (#6306)

This reverts commit 9df9f92.

Clear out the start_ts field in the postings of deltas before they are
marshalled to avoid storing that field in disk. This field is only meant
to be used during in-memory processing.

Related to https://discuss.dgraph.io/t/start-ts-not-being-cleared-before-postings-are-written-to-disk/9146

(cherry picked from commit fbbd731)

Reference: [Discuss Post](https://discuss.dgraph.io/t/custom-dql-typename-bug-strange-behaviour/9341)
(cherry picked from commit 0faae40)

Fixes: #5516
Fixes: [Discuss Issue](https://discuss.dgraph.io/t/graphql-fragments-generates-unexpected-behaviour/7470)
(cherry picked from commit dcce8bf)

Fixes DGRAPH-2181

The queries in dgraph are processed in separate goroutines. The pstore badger DB
could be closed while the query was being processed. This causes panics such as
```
panic: runtime error: invalid memory address or nil pointer dereference
	panic: Unclosed iterator at time of Txn.Discard.
[signal SIGSEGV: segmentation violation code=0x1 addr=0x10 pc=0x11d2311]

goroutine 19298 [running]:
github.com/dgraph-io/badger/v2.(*Txn).Discard(0xc05586bc20)
	/go/pkg/mod/github.com/dgraph-io/badger/v2@v2.0.1-rc1.0.20200718033852-37ee16d8ad1c/txn.go:517 +0xc1
panic(0x19e26a0, 0x2988560)
	/usr/local/go/src/runtime/panic.go:969 +0x166
github.com/dgraph-io/badger/v2/skl.(*Skiplist).IncrRef(...)
```
This PR attempts to reduce the number of such crashes. This PR doesn't fix the actual
issue but it tries to reduce the probability of such crashes by checking if badger is not
closed before accessing it.

An ideal fix would be to stop all the goroutines started by dgraph while closing DB so
that we don't try to read from a closed badger DB.

(cherry picked from commit 3cea0fe)

Backups should be added to the list of tasks (rollups, snapshots, etc)
that are managed by Dgraph so that rollups and other tasks are paused
during the backup.

(cherry picked from commit ea15b66)

This badger update brings the following changes from badger.

https://github.com/dgraph-io/badger/commits/b41e77a Fix(cleanup): Avoid truncating in value.Open on error
https://github.com/dgraph-io/badger/commits/a0d4903 fix(cleanup): Do not close cache before compaction

(cherry picked from commit deded8b)

(cherry picked from commit a8a6e85)

Co-authored-by: Tejas Dinkar <tejas@gja.in>

…quest at a time. (#6339)

It's possible that two requests reach the server around the same time
and send a requests to the alphas with the same backupNum. This could
lead to issues further down the line.

Related to DGRAPH-2295

(cherry picked from commit 5b79260)

… mode for zero. (#6342)

Related to DGRAPH-2189

(cherry picked from commit bf79999)

(cherry picked from commit 844f34d)

Fixes DGRAPH-2357.
Fixes GRAPHQL-562.

(cherry picked from commit 91e28d5)

…#6352)

Fixes DGRAPH-2356
This PR fixes random GraphQL custom logic e2e test failures in TeamCity by requiring docker-compose to always build images before starting the containers, as the mock server in custom logic tests can change any time.

(cherry picked from commit 536b317)

…GQLSchema (#6349) (#6354)

* fix(GraphQL): incorrect generatedSchema in updateGQLSchema (#6349)

This PR fixes the behaviour where Dgraph schema was being given as generatedSchema field in updateGQLSchema request. Now, the newly generated complete GraphQL schema is given as generatedSchema.

(cherry picked from commit 0b8681c)

# Conflicts:
#	graphql/e2e/schema/schema_test.go

…#6355)

(cherry picked from commit 254bd29)

(cherry picked from commit 7857fa2)

remove duplicate word "doesn't" from description of two-way edges.

(cherry picked from commit e165710)

The closer type was moved from badger/y to ristretto/z in dgraph-io/ristretto#191

(cherry picked from commit 0287838)

(cherry picked from commit 0fab710)

…6402)

* Fix(Alpha): MASA: Make Alpha Shutdown Again (#6313)

This PR removes the usage of context.Background() in groups.go and ensures that all the various goroutines exit as intended.

Changes:
* Shutdown SubscribeForUpdates correctly.
* Fix up all the closing conditions.
* Consolidate updaters into one closer
* Update Badger to master
* fix(build): Update ResetAcl args for OSS build.
* chore: Remove TODO comment.

Co-authored-by: Daniel Mai <daniel@dgraph.io>
(cherry picked from commit f1941b3)

…6359) (#6407)

Forcefully kill alpha if CTRL-C is caught before node initialization completes.

Fixes DGRAPH-2377

Co-authored-by: Manish R Jain <manish@dgraph.io>
(cherry picked from commit 41c2052)

… (#6336)

(cherry picked from commit d94e5da)

…ommunity. (#6422)

Fixes DGRAPH-2317. Issue is with v20.07 we restrict the admin operations from alpha instances only. But this hasn't been mentioned during the deployment process in any examples. Hence community is running into issues like Unauthorized IP address X.X.X.X. This doc addition in troubleshoot page will help the community members if they run into this issue.

The different generated proto file makes the release.sh script fail
with "Generated protos different in release."

Changes
* Called "make regenerate"

Fixes GRAPHQL-673.
This PR fixes the issue where introspection queries would break if there were two types implementing same interface and having fields with same name in those two types and that repeating field is also a field in introspection query.
This was introduced after #6228.

(cherry picked from commit 243a336)

) (#6413)

Fixes GRAPHQL-650.

For this user schema:
```
type Starship {
        id: ID!
        name: String! @search(by: [term])
        length: Int64
        tags: [String]
        createdAt: DateTime
}
```
we were generating complete GraphQL schema with StarshipOrderable as:
```
enum StarshipOrderable {
  name
  length
  tags
  createdAt
}
```
that would cause a DQL query to be formed which errors out, see this GraphQL query:
```
query {
  queryStarship(order: {asc: tags}) {
    id
    name
    length
    tags
  }
}
```
It gives back:
```
{
  "errors": [
    {
      "message": "Dgraph query failed because Dgraph execution failed because : Sorting not supported on attr: Starship.tags of type: [scalar]"
    }
  ],
  "data": {
    "queryStarship": []
  }
}
```
which means we should not allow even list of scalar along with object types in orderable. Only scalar field should be allowed in orderable. So, the correct orderable should be without tags field:
```
enum StarshipOrderable {
  name
  length
  createdAt
}
```

This PR fixes the above bug.

(cherry picked from commit dc66617)

#6421)

* Add separate compression flag for z and wal dirs

* Address comments

* Address comments

(cherry picked from commit 601cc3b)

Fixes GRAPHQL-657
Fixes [Discuss Issue](https://discuss.dgraph.io/t/documentation-that-arrays-work-as-a-set-not-storing-duplicates/9590)

(cherry picked from commit 07cff6f)

…erable field in a ty… (#6460)

fix(GraphQL): Generate correct schema when no orderable field in a type (#6456)

Fixes GRAPHQL-650
It was introduced after #6392.

(cherry picked from commit 8e4d121)

(cherry picked from commit 512c69b)

…ingle UID predicate in the same transaction. (#6449)

The iterate method does not account for delete postings in the same transaction
so they have to be preserved manually. Otherwise, if a deletion and an add operation
to the same uid happen in the same predicate, the deletion operation is not
applied.

Fixes DGRAPH-1309

(cherry picked from commit 56c8f3c)

* Fix references in docs

* Fix mutations/language-rdf-types.md reference

(cherry picked from commit 027db53)

Use the readTs for reads via LocalCache. This would ensure that users don't get a lot of ReadTs less than MinTs errors. This PR also fixes the flaky test in increment_test.go.

This has a side-effect on posting list cache. That cache would need to understand versioning to be effective.

Co-authored-by: Daniel Mai <daniel@dgraph.io>
(cherry picked from commit 2de4675)

Co-authored-by: Manish R Jain <manish@dgraph.io>

* Fix order and offset in auth queries.

* Remove order from top-level auth filter.

(cherry picked from commit a5bfd4a)

… with inverse field. (#6350) (#6524)

* Fix query rewriting for auth delete when deleting types with inverse field.

(cherry picked from commit be9ebd0)

…licate xids within t… (#6521)

Fixes GRAPHQL-641

See https://discuss.dgraph.io/t/residual-issue-linking-grandchild-in-mutation-with-custom-ids/9613 for the issue details.

If the same nested xid was used as a leaf node within an addType mutation, it was only linked the first time that it was used and not subsequently for later objects. This change fixes that by adding the linkage mutations and only removing the mutation which was creating the object with the xid again.

(cherry picked from commit 12b2716)

# Conflicts:
#	graphql/e2e/directives/dgraph_directives_test.go
#	graphql/e2e/normal/normal_test.go
#	graphql/resolve/add_mutation_test.yaml

…ion after drop all (#6268) (#6525)

This PR fixes the "Internal error" response when querying the GraphQL schema after performing drop_all operation.

(cherry picked from commit d3bee33)

) (#6417)

Fixes GRAPHQL-655.

For this user schema (interface have no non-Id field):

```
interface A {
   name: String! @id
}

type B implements A {
  age: Int!
}
```
the following delete mutation
```
mutation{
  deleteA(filter:{name:{eq: "xyz"}}){
    a{
      name
    }
  }
}
```
was resulting in the following error:
```
{
  "errors": [
    {
      "message": "Internal Server Error - a panic was trapped.  This indicates a bug in the GraphQL server.  A stack trace was logged.  Please let us know by filing an issue with the stack trace."
    }
  ]
}
```
This PR fixes this bug and now deleting can be performed successfully.

(cherry picked from commit 742259b)

…6424)

Fixes GRAPHQL-642.

For this restoreStatus query using variable
```
query restoreStatus($restoreId: Int!) {
	restoreStatus(restoreId: $restoreId) {
		status
		errors
	}
}
```
was giving this panic
```
panic: interface conversion: interface {} is json.Number, not int64.
```
Whereas the expected result should be
```
{
  "data": {
    "restoreStatus": {
      "status": "UNKNOWN",
      "errors": []
    }
  },
  "extensions": {}
}
```
This PR fixes this panic, Now `resolveStatus` with or without variable works fine.

(cherry picked from commit 45afae9)

…6330) (#6522)

This PR fix errors from authorization examples given in docs.

(cherry picked from commit 0f3bfb1)

)

(cherry picked from commit 5aca255)

* Fix(GraphQL): This PR add schema cleaning in GraphQL. (#6427)

Fixes GRAPHQL-572

This PR adds schema cleaning in GraphQL which will remove empty links and types from generated GraphQL schema recursively.

(cherry picked from commit bec5322)

…ed (#6416) (#6530)

While creating the add mutation, we weren't squashing different bits correctly which led to the incorrect mutation being sent to Dgraph. This change modifies the set to append for []interface{} to fix that.

Fixes GRAPHQL-679

(cherry picked from commit 816a08f)

# Conflicts:
#	graphql/e2e/common/mutation.go
#	graphql/e2e/directives/schema.graphql
#	graphql/e2e/normal/schema.graphql
#	graphql/resolve/schema.graphql

…6535)

* fix(GraphQL): Fix cascade with auth query when RBAC is false (#6444)

* Fix cascade with auth query.

* Added additional E2E test.

(cherry picked from commit 8c518d4)

…h rule. (#6391) (#6534)

* fix(GraphQL): Hide info when performing mutation on id field with auth rule. (#6391)

* Hide info when performing mutation on id field with auth rule.

(cherry picked from commit 5c33428)

…6373) (#6554)

Fixes GRAPHQL-631
This PR adds documentation for Scalar DateTime type.

(cherry picked from commit 91c81ca)

#6559)

This reverts commit 812a6f0.

# Conflicts:
#	graphql/schema/testdata/schemagen/output/filter-cleanSchema-all-empty.graphql
#	graphql/schema/testdata/schemagen/output/filter-cleanSchema-circular.graphql
#	graphql/schema/testdata/schemagen/output/filter-cleanSchema-custom-mutation.graphql
#	graphql/schema/testdata/schemagen/output/filter-cleanSchema-directLink.graphql

(cherry picked from commit 512c0a9)

This PR changes how we use badger in the wal store. Currently, we run
badger in normal mode for `w` and `zw` store. We've seen up to 900K
entries for the same hard state (hs) key in `zw` store. These duplicate
keys cause spikes in read latencies in `zw`. The `w` store has more
compactions compared to `zw` store and so it has lesser stale data and
thus lesser spikes in read latencies.

The fix here is to open the `w` and `zw` directories in managed mode and
perform all writes on the same timestamp (max version in the db). This
leads to close to 0 duplicates in the store.

This PR also fixes the raft leader election issue which is a result of high
read latencies.

(cherry picked from commit 6882e37)

This PR adds `Content-Length` header to the response of queries sent via HTTP.

Fixes DGRAPH-1675

(cherry picked from commit ed63786)

Bulk loader uses caches in compression and this PR adds flags to make it
configurable.

Bulk loader was setting compressionLevel but not the compression option.
As a result of this, badger wasn't compressing any data. This PR fixes
this setting compression if compressionLevel is greater than 0.

(cherry picked from commit 99341dc)

…6590)

* (release/v20.07): feat(Dgraph): add utility to export backup data. (#6550)

This utility allows to take a single backup (full or incremental) and export the data
and the schema inside it to RDF. Encrypted backups are supported.

Fixes DGRAPH-2465

(cherry picked from commit 369a5c1)

* Remove destination field.

…ner. (#6601)

The script is useful for situations in which the host system creates a
binary that's incompatible with the docker container (e.g different
versions of glibc).

(cherry picked from commit c789738)

(cherry picked from commit b777913)

The test and utility were assuming that node 1 was in group 1, node 2 in
group 2, and so on. This is not true all the time, hence the flakiness.

Fixing the code to not make this assumptions as well as retrying the
DropAll in NodesSetup to ensure all the nodes are up before proceeding
with the test.

(cherry picked from commit 36638a7)

(cherry picked from commit 09ef988)

Enterprise License unit tests

#6634)

fixed : Pagination param "after" does not work when using func: uid(v)

… > 0. (#6597)


(cherry picked from commit a4f21f4)

* fix(gqlgen): Update gqlgen in go.mod (#6594)

* Add tests for querying alias and attribute on interface

* Requested changes

(cherry picked from commit f19d882)

* Fix test

This PR fixes incorrect linking of objects with `hasInverse`. For example
For this schema, `type Country` has field `state` with `hasInverse` predicate.
```
type Country {
        id: ID!
        name: String! @search(by: [trigram, hash])
        states: [State] @hasInverse(field: country)
}

type State {
        id: ID!
        xcode: String! @id @search(by: [regexp])
        name: String!
	capital: String
	country: Country
}
```
And for the following mutation
```
mutation addCountry($input: [AddCountryInput!]!) {
			addCountry(input: $input) {
			  country {
				name
				states{
				  xcode
				  name
				  country{
					name
				  }
				}
			  }
			}
		  }
```
with `input`:
```
{
	"input": {
		"name": "A Country",
		"states": [
			{
				"xcode": "abc",
				"name": "Alphabet"
			},
			{
				"xcode": "def",
				"name": "Vowel",
				"country": {
					"name": "B country"
				}
			}
		]
	}
}
```
should result in
```
{
		"addCountry": {
		  "country": [
			{
			  "name": "A country",
			  "states": [
				{
				  "country": {
					"name": "A country"
				  },
				  "name": "Alphabet",
				  "xcode": "abc"
				},
				{
				  "country": {
					"name": "A country"
				  },
				  "name": "Vowel",
				  "xcode": "def"
				}
			  ]
			}
		  ]
		}
	  }`
```
whereas the incorrect output was
```
{
		"addCountry": {
		  "country": [
			{
			  "name": "A country",
			  "states": [
				{
				  "country": {
					"name": "A country"
				  },
				  "name": "Alphabet",
				  "xcode": "abc"
				},
				{
				  "country": {
					"name": "B country"
				  },
				  "name": "Vowel",
				  "xcode": "def"
				}
			  ]
			}
		  ]
		}
	  }`
```
This PR fixes this issue.

(cherry picked from commit e496467)

This PR updates documentation for the `/admin` endpoint operation and  `restore` operation.

(cherry picked from commit 288eafb)

…) (#6647)

(cherry picked from commit cec5567)
This PR fixes: GRAPHQL-638

This is related to GRAPHQL-564. as is Dgraph reserved keyword. It's being used by DQL to identify variables. This PR is to restrict the naming for fields with the name as.

(cherry picked from commit 014ee10)

The k-shortest path query sometimes returns a wrong response. The cause
of this issue is inappropriate use of sync.pools. Whenever a valid path
is found a new variable is assigned with this route and that variable is
appended in the kroutes (which is then parsed to produce response).
But the route is also put in the sync.pool. Since the route is a
structure containing a pointer to slice along with other fields, when
it is fetched back from the pool in future then any mutation on this fetched
route causes modification in the kroutes, resulting in garbage response.

This issue is fixed by making a copy of the route struct rather
than just assigning it to a variable, which causes a shallow copy.

(cherry picked from commit 4792d8b)

(cherry picked from commit a8a47af)

Co-authored-by: Manish R Jain <manish@dgraph.io>

Fixes DGRAPH-2419
Fixes [Discuss Issue](https://discuss.dgraph.io/t/acl-login-will-fail-if-auth-token-enabled-in-v20-07-0/10044)

This PR fixes Poor-man's auth for following endpoints:
* `/login`
* `/admin`

(cherry picked from commit 4fc328d)

# Conflicts:
#	graphql/e2e/common/common.go

…imanyu/release-slash-merge

# Conflicts:
#	dgraph/cmd/alpha/run.go
#	dgraph/cmd/bulk/reduce.go
#	dgraph/cmd/zero/raft.go
#	edgraph/server.go
#	go.mod
#	go.sum
#	graphql/e2e/common/common.go
#	graphql/e2e/common/fragment.go
#	graphql/e2e/common/mutation.go
#	graphql/e2e/common/query.go
#	graphql/e2e/custom_logic/custom_logic_test.go
#	graphql/e2e/directives/schema_response.json
#	graphql/e2e/normal/schema_response.json
#	graphql/e2e/schema/generatedSchema.graphql
#	graphql/resolve/resolver.go
#	graphql/schema/testdata/schemagen/output/type-without-orderables.graphql
#	protos/pb/pb.pb.go
#	wiki/content/enterprise-features/binary-backups.md
#	wiki/content/graphql/admin/index.md
#	wiki/content/graphql/schema/reserved.md

…imanyu/release-slash-merge

# Conflicts:
#	dgraph/cmd/alpha/run.go
#	dgraph/cmd/bulk/reduce.go
#	dgraph/cmd/zero/raft.go
#	edgraph/server.go
#	go.mod
#	go.sum
#	graphql/e2e/common/common.go
#	graphql/e2e/common/fragment.go
#	graphql/e2e/common/mutation.go
#	graphql/e2e/common/query.go
#	graphql/e2e/custom_logic/custom_logic_test.go
#	graphql/e2e/directives/schema_response.json
#	graphql/e2e/normal/schema_response.json
#	graphql/e2e/schema/generatedSchema.graphql
#	graphql/resolve/resolver.go
#	graphql/schema/testdata/schemagen/output/type-without-orderables.graphql
#	protos/pb/pb.pb.go
#	wiki/content/enterprise-features/binary-backups.md
#	wiki/content/graphql/admin/index.md
#	wiki/content/graphql/schema/reserved.md