Was this page helpful?
Query execution time can be limited by setting a request timeout. If a query does not complete
in the given time, then QueryError::RequestTimeout is returned by the driver immediately,
so that application logic can continue operating, but the query may still be in progress on the server.
As a side note, if one wishes custom server-side timeouts (i.e. actual interruption of query processing),
one can use aUSING TIMEOUT <duration> directive supported in ScyllaDB
(but not in Cassandra).
Timeout can be set globally (per session) or locally (for given statement).
The default per-session timeout is currently 30s.
It is possible to turn off timeouts completely by providing None as timeout when building Session.
However, setting per-statement timeout to None results in falling back to per-session timeout.
use scylla::{Session, SessionBuilder, query::Query};
use scylla::transport::ExecutionProfile;
use std::time::Duration;
let uri = std::env::var("SCYLLA_URI")
.unwrap_or_else(|_| "127.0.0.1:9042".to_string());
let no_timeout_profile_handle = ExecutionProfile::builder()
.request_timeout(None) // no timeout
.build()
.into_handle();
let session: Session = SessionBuilder::new()
.known_node(uri)
.default_execution_profile_handle(no_timeout_profile_handle) // no per-session timeout
.build()
.await?;
// This query, having no timeout, could block indefinitely if a queried node hangs.
session
.query("TRUNCATE keyspace.table", ())
.await?;
let three_sec_timeout_profile_handle = ExecutionProfile::builder()
.request_timeout(Some(Duration::from_secs(3))) // no timeout
.build()
.into_handle();
// The below query will last for no more than 3 seconds, yielding a RequestTimeout error
// if no response arrives until then.
let mut query: Query = "TRUNCATE keyspace.table".into();
query.set_execution_profile_handle(Some(three_sec_timeout_profile_handle));
session
.query(query, ())
.await?;
Was this page helpful?