Arango restarts on a random basis when same traversal query is performed

Question

I am running a series of simultaneous requests from a node server to an Ubuntu AWS Instance with Arango installed. Arangod is version 3.1.7

Everything worked fine for a few weeks, but now :

On a random basis, I get an {Error: socket hang up} from node while performing on of my queries, always the same one, and arango seems to restart as I get {Error: connect ECONNREFUSED} on all other queries for a little while. It happens 1 out of 6 times, although my data did not change between queries.

Web Interface logs are very few and don't seem to explain what went wrong. Arangod seems to create a logfile, but it is very hard to interpret. All server metrics are fine.

Do you have ideas on how can I move forward on this issue ? Is my data corrupted ? How can I find out ?

Here is the content of my logfiles :

2017-02-22T16:24:57Z [11449] INFO ArangoDB 3.1.7 [linux] 64bit, using VPack 0.1.30, ICU 54.1, V8 5.0.71.39, OpenSSL 1.0.1f 6 Jan 2014
2017-02-22T16:24:57Z [11449] INFO using SSL options: SSL_OP_CIPHER_SERVER_PREFERENCE, SSL_OP_TLS_ROLLBACK_BUG
2017-02-22T16:24:57Z [11449] INFO Authentication is turned on
2017-02-22T16:24:57Z [11449] INFO Authentication system only
2017-02-22T16:24:57Z [11449] INFO Authentication for unix sockets is turned on
2017-02-22T16:24:57Z [11449] INFO file-descriptors (nofiles) hard limit is 131072, soft limit is 131072
2017-02-22T16:24:57Z [11449] INFO running WAL recovery (1 logfiles)
2017-02-22T16:24:57Z [11449] INFO replaying WAL logfile '/vol/data/standalone/journals/logfile-44002553.db' (1 of 1)
2017-02-22T16:24:57Z [11449] INFO WAL recovery finished successfully
2017-02-22T16:24:57Z [11449] INFO JavaScript using startup '/usr/share/arangodb3/js', application '/var/lib/arangodb3-apps'
2017-02-22T16:24:57Z [11449] INFO In database '_system': Database is up-to-date (30107/standalone/existing)
2017-02-22T16:24:57Z [11449] INFO In database 'X-DB': Database is up-to-date (30107/standalone/existing)
2017-02-22T16:24:57Z [11449] INFO using endpoint 'http+tcp://0.0.0.0:8529' for non-encrypted requests
2017-02-22T16:24:58Z [11449] INFO Please note that a new bugfix version '3.1.11' is available
2017-02-22T16:24:58Z [11449] INFO ArangoDB (version 3.1.7 [linux]) is ready for business. Have fun!

And :

2017-02-22T16:24:57Z [1311] ERROR {startup} child 11302 died a horrible death, signal 11
2017-02-22T16:27:57Z [1311] ERROR {startup} child 11449 died a horrible death, signal 11

I believe the query causing the trouble is :

RETURN (
FOR i, j, k IN 0..3 INBOUND "Hint/26427214" GRAPH "X-graph"
FILTER k.vertices[1]._id LIKE "Drop/%"
AND k.vertices[2]._id LIKE "Hint/%"
AND k.vertices[3]._id LIKE "Drop/%"
FILTER k.vertices[1].hostname != k.vertices[3].hostname
FILTER k.vertices[2].type == "article"
SORT k.vertices[2].createdAt DESC
LIMIT 0, 1
RETURN UNION (
FOR v, e, p IN 0..2 INBOUND k.vertices[2]._id GRAPH "X-graph"
FILTER p.vertices[1]._id LIKE "Source/%"
AND p.vertices[2]._id LIKE "Twittos/%"
RETURN { key: p.vertices[0]._key, retweets: p.vertices[0].retweets, title: p.vertices[0].title, type: p.vertices[0].type, content: p.vertices[0].content, image: p.vertices[0].image, source: { key: p.vertices[1]._key, name: p.vertices[1].name, screenname: p.vertices[2].screenname, image: p.vertices[2].image }},
FOR v, e, p IN 0..1 INBOUND k.vertices[2]._id GRAPH "X-graph"
FILTER p.vertices[1]._id LIKE "Source/%"
RETURN { key: p.vertices[0]._key, retweets: p.vertices[0].retweets, title: p.vertices[0].title, type: p.vertices[0].type, content: p.vertices[0].content, image: p.vertices[0].image, source: { key: p.vertices[1]._key, name: p.vertices[1].name }},
FOR v, e, p IN 0 INBOUND k.vertices[2]._id GRAPH "X-graph" RETURN { key: p.vertices[0]._key, retweets: p.vertices[0].retweets, title: p.vertices[0].title, type: p.vertices[0].type, image: p.vertices[0].image, content: p.vertices[0].content })[0] )[*]

Answer 1

After a lot of research, I tried to simply upgrade to arangod 3.1.11. The issue seems to be gone.