Oracle 12.1 on AIX
We started getting ORA-55917 in alert.og:
Thread 1 advanced to log sequence 99075 (LGWR switch)
Current log# 3 seq# 99075 mem# 0: /db01/oracle/redo_g03_m01.log
Mon Apr 01 20:17:41 2019
Archived Log entry 99050 added for thread 1 sequence 99074 ID 0xffffffffc8b2f32c dest 1:
Mon Apr 01 20:38:55 2019
opidcl aborting process unknown ospid (5833820) as a result of ORA-55917
Mon Apr 01 20:41:59 2019
Thread 1 advanced to log sequence 99076 (LGWR switch)
Current log# 4 seq# 99076 mem# 0: /db01/oracle/redo_g04_m01.log
This is datawarehouse-type database with complex queries based mostly on full table scans/parallel query and hash joins.
Very little information is available about this error:
https://docs.oracle.com/database/121/ERRMG/ORA-54500.htm
ORA-55917: Table flush I/O failed for log ID: string bucket ID: string
Cause: Table flush I/O failed. The most common cause for this could be out-of-space errors in the tablespace.
This explanation doesn't really help. What exactly is 'table flush'? As far as I know Oracle doesn't flush tables, db writer process writes updated ("dirty") blocks to disk.
In this database the main tablespace is 20 TB in size, it is 73% full, so more than 5 TB is free. There are no errors in alert.log indicating that a tablespace was out of space.