0

I have setup a two node MariaDB Galera cluster on Ubuntu systems. A simple application connects to a database using MaxScale and it works fine. But when the node in the cluster that is currently in use, say, node 1, fails, the application gets error such as 1927 or 1045. On receiving this error, the application tries to connect to database again but it keeps failing many times but succeeds once fail over from node 1 to node 2 is complete and MaxScale gives database connection to node 2. The connection trial duration ranges from 20 to 50 seconds in my cluster environment.

My question is whether or not there is any MaxScale connection time out parameter that I can use to specify connection timeout to some value such as 50 seconds so that application tries just once for a new connection instead of trying many times. (I used parameter connectTimeout in JDBC URL for the database but it was not effective for my application and I think this is expected.)

mvsagar
  • 1,998
  • 2
  • 17
  • 19

1 Answers1

2

MaxScale is sending errors most likely because no master server is available. This error cannot be prevented with MaxScale 2.2 and client side re-connection is required.

In MaxScale 2.3, a new feature will be available that allows similar behavior to what you describe (see MXS-1501).

If you are performing read-only requests, it might be beneficial to enable master_failure_mode=error_on_write. This will allow read-only requests to be done even when no master server is available.

markusjm
  • 2,358
  • 1
  • 11
  • 23
  • Yes, no master status was displayed by maxadmin during connection retrial period but status of second node changed to master at the end of the period and probably at this time max scale could give valid connection to the second node. Thank you for your reply and letting me know required feature would be available in future version of MaxScale. – mvsagar May 28 '18 at 07:57