Perhaps we can cache the timed out queries, so we don't need to "explain" them more than once.
On Wed, Mar 15, 2017 at 11:07 AM, michael.griggs < [email protected]> wrote: > In IgniteH2Indexing.java#executeSqlQueryWithTimer we automatically issue a > second query to "EXPLAIN" why a query has timed out. I think we should > make > optional the automatic issuance of a second "EXPLAIN" query. When a query > times out it could very well be due to high load on the system, and issuing > a second query for every timed-out one will only increase system load, > leading to more timeouts, and so on... > ResultSet rs = executeSqlQuery(conn, stmt, timeoutMillis, > cancel); long time = U.currentTimeMillis() - start; > long longQryExecTimeout = > schemas.get(schema(space)).ccfg.getLongQueryWarningTimeout(); > if > (time > longQryExecTimeout) { String msg = "Query execution > is too long (" + time + " ms): " + sql; ResultSet plan = > executeSqlQuery(conn, preparedStatementWithParams(conn, "EXPLAIN " + sql, > params, false), 0, null); > What do others think? > > > > -- > View this message in context: http://apache-ignite- > developers.2346864.n4.nabble.com/Timeout-on-a-SQL-query- > issues-another-SQL-query-to-generate-EXPLAIN-tp15475.html > Sent from the Apache Ignite Developers mailing list archive at Nabble.com.
