The log shows:
com.graphdbapi.driver.v1.exceptions.DatabaseException:Cypher task execution blocking time overruns
What causes this problem? How to solve it?
I feel like it’s because there are too many concurrent queries. Is concurrency not possible? When I was monitoring, I noticed that during certain periods of high queries, disk reads are much higher than disk IO, why is that?
>Solution :
Yes, there are too many Cypher queries in the queue that haven’t finished yet, and new Cypher requests have timed out.
Concurrency is limited and can be modified in the ‘User Management’ section on the page 8888. Each query incurs resource costs, and when concurrency is high, resources become scarce, increasing the overhead of resource switching.
According to your comment on this answer: The node monitoring on page 51314 of Galaxybase uses the Java OSHI library. Both the Java OSHI library and the ‘iostat’ command in the server can obtain disk information, but the data range and granularity obtained may slightly differ. The Java OSHI library monitors the IO of the entire server, while ‘iostat’ can provide specific details about a particular device or data disk. Therefore, when the IO speed displayed on the Galaxybase monitoring page is higher than the IO speed obtained from the ‘iostat’ command in the server, you need to make sure if the server is solely dedicated to serving the graph database. In a multi-node cluster, the monitoring page of Galaxybase also monitors the disk read rate of each node, using the same method. It is important to check if the server hosting the abnormal node has any other IO tasks.