When you see something like - “the configuration database is locked by session none” - it is typically an internal “job” that is holding a lock on CDB. For example:
- You use HA and an out of sync/new standby node connects to the active node (aka slave/master) it will trigger a sync.
- You have a CDB subscriber that’s slow to respond in some cases.
- You have a southbound client (CDB, MAAPI etc) that for some reason holds a lock on CDB.
You should not see a crash when a transaction is aborted, but you can try to increase the confdConfig/commitRetryTimeout in confd.conf for debug purposes to try to find what cause the delay, especially if you have triggered a deadlock
For example, if you try with commitRetryTimeout PT600S and you still get a “the configuration database is locked by session none” response then there could be a deadlock due to, for example, a CDB subscriber application triggering itself by committing something. If you only increase the commitRetryTimeout to PT10S and you no longer observe the issue, you check the developer log (developerLogLevel trace) for more detailed information. Also, as suggested by @mvf , the errorLog will likely be helpful.