Alpha crashes with fault address in groupby.go

[jarifibrahim]

What version of Dgraph are you using?

727b246

Have you tried reproducing the issue with latest release?

Yes

What is the hardware spec (RAM, OS)?

N/A

Steps to reproduce the issue (command/config used to run Dgraph).

1. Run zero using

dgraph zero -w /home/ibrahim/m2_drive/flock_backup/zero1/zw -o 0 --idx=1 --my=localhost:5080 --replicas=1 --logtostderr -v=2 --bindall

2. Run alpha using

dgraph alpha -p /home/ibrahim/m2_drive/flock_backup/alpha1/p -w /home/ibrahim/m2_drive/flock_backup/alpha1/w -o 100 --my=localhost:7180 --lru_mb=1024 --zero=localhost:5080 --logtostderr -v=2

4. Apply the following diff on flock agent's main.go

diff --git a/main.go b/main.go
index c2cf6ec..91e8da0 100644
--- a/main.go
+++ b/main.go
@@ -52,8 +52,8 @@ id_str: string @index(exact) @upsert .
 created_at: dateTime @index(hour) .
 hashtags: [string] @index(exact) .
 
-author: uid @count @reverse .
-mention: uid @reverse .
+author: [uid] @count @reverse .
+mention: [uid] @reverse .
 `
 
        cDgraphTweetQuery = `

3. Run flock agent

go run main.go -a ":9180"

4. Apply the following to flock client's main.go (without this change some of the queries in flock were failing)

diff --git a/client/main.go b/client/main.go
index c1eab19..5913475 100644
--- a/client/main.go
+++ b/client/main.go
@@ -547,7 +547,7 @@ type querySix struct {
 
 func (q *querySix) getParams(dgr *dgo.Dgraph) error {
        // we subtract 41 hours because that's the latest data we get from twitter
-       curTime := time.Now().Add(-41 * time.Hour)
+       curTime := time.Now().Add(-100 * time.Hour)
 
        query := fmt.Sprintf(`
 {
@@ -619,7 +619,7 @@ type querySeven struct {
 
 func (q *querySeven) getParams(dgr *dgo.Dgraph) error {
        // we subtract 41 hours because that's the latest data we get from twitter
-       curTime := time.Now().Add(-41 * time.Hour)
+       curTime := time.Now().Add(-100 * time.Hour)
 
        query := fmt.Sprintf(`
 {
@@ -693,7 +693,7 @@ type queryEight struct {
 
 func (q *queryEight) runQuery(dgr *dgo.Dgraph) error {
        // we subtract 41 hours because that's the latest data we get from twitter
-       curTime := time.Now().Add(-41 * time.Hour)
+       curTime := time.Now().Add(-100 * time.Hour)
 
        query := fmt.Sprintf(`
 {
@@ -787,7 +787,7 @@ type queryNine struct {
 
 func (q *queryNine) getParams(dgr *dgo.Dgraph) error {
        // we subtract 41 hours because that's the latest data we get from twitter
-       curTime := time.Now().Add(-41 * time.Hour)
+       curTime := time.Now().Add(-100 * time.Hour)
 
        query := fmt.Sprintf(`
 {

5. Run flock client

go run main.go -a ":9180"

Expected behaviour and actual result.

After a few seconds alpha crashes. Here's the stack trace of alpha https://gist.github.com/jarifibrahim/d517efefd7f7c334a9bb50552041cc11

[martinmr]

@jarifibrahim Can you check if the following change to groupby.go in line 225 helps?

Change from

ul := child.uidMatrix[i]

to

if i >= len(child.uidMatrix) {
  continue
}
ul := child.uidMatrix[i]

I think that would stop the crash but I am not sure if it's the right solution, or why uidMatrix does not have enough entries.

[jarifibrahim]

@martinmr The change stopped the crash.

[martinmr]

ok.still not sure of the root cause but i'll open a PR with the fix. We have similar checks in our codebase so this should be fine.

[martinmr]

My bad, the crash is in the next line while trying to access ul.Uids. So ul is nil. I am guessing the error is not happening every time so it looked like my original suggestion fixed it. For now, I'll open a PR to check that ul is non-nil.

[martinmr]

#3670

[martinmr]

Fix has been submitted so closing this for now. Feel free to reopen if you happen to see it again but the original segfault should be gone now.