After this, we tried connecting once every client run (which is in the order of minutes), run all queries with that connection and then explicitly disconnecting.After this, we tried keeping a single connection alive for all subsquent connections.First we were using a single conenction per query, which meant connecting and querying every 500 ms - 10 seconds.We are using the latest pyArango library (v2.0.2), and have tried different connection strategies. Only by reseting the machine can I bring the system back up again. This even makes the host machine unresponsive some times. Everything seems fine until after some days I start seeing a ConnectionResetError(104, 'Connection reset by peer')) on the clients without any correlation to any particular query, It can happen in after trying any query.Īfter this error occurs, most of the times the clients can no longer connect to arango, which refuses al connections. I have a setup where 5-8 containers run constant queries on the arango container. Error happens after running system for a few days.
Replication Factor & Number of Shards (Cluster only): Steps to reproduce FOR doc IN col FILTER doc.`_key` = mykey RETURN doc._keyįOR doc IN col FILTER doc.`id` = myid AND doc.`date` >= AND doc.`date` <= RETURN doc._keyĪQL explain and/or profile (if applicable):ĭatasets are usually small dictionaries ( 5-15 fields, and one or two fields having from 1 - 200 elements).