Hi,
We are experiencing connection issues between our application on Azure VM 1 and Neo4j on Azure VM 2. After a long time of troubleshooting I am out of ideas what could be the underlying issue.
Neo4j version: 4.3.2
VM: Azure
Installation: Windows service
What is happening?
Several times through-out the day our application shoots a Cypher call to the Neo4j Bolt port. At this moment 1 call per minute, but this could even be 1 call per 5/10/15 minutes. Most of the times Neo4j responds very fast (<1 s) and we can render the response in our application. However, there are moments that Neo4j takes more than 5 seconds to respond. Even with the same query. Sidenote: the query is very simple and returns a maximum of 150 records within 5 ms. After some time tracking this problem, it seems that there is more change that these issues occur when the application does not shoot a Neo4j call in a long time. So, most of the time these issues happen early on in the morning.
What did we already tried?
- Investigating the Neo4j log files. It seems that Neo4j almost instantly receives the Cypher call, so it seems that there is no connection issue between the Azure VM's.
- We thought that maybe Neo4j is in a sort of 'sleep mode' or something. Thus, we shoot 'simple warm up'-calls if the user is in a flow to activate Neo4j so later-on the time-out does not occur. But this does not seem to fix this issue.
- We installed Sentry.io to track our application whether there are issues on the processing side and not Neo4j, but this is also not the case.
- (We still want to investigate if this issue disappears by upgrading Neo4j version 4.3.2 to version 4.4.6.)
So I am bit left in the dark what is going on. Especially since the connection time-out seem to happen randomly. I still feel that this issue that we are experiencing is happening through Neo4j. Does anybody recognize this issue or has an idea what could be the underlying issue?
Thanks in advance 🙂
