enq: tt应当与 回收recyclebin中对象空间 以及segment分配extent有关
TOP SQL是:
insert /*+ append */ into t_indoor_leak_sampling
insert append大量数据是会引起回收recyclebin中对象空间 以及segment分配extent的 。
但为什么enq: TT - contention会这么慢 ,可能与IO有关,至少目前看到的Io并不好
db file sequential read 2,368,962 13,837 6 19.47 User I/O
log file sync 196,804 2,322 12 3.27 Commit
db file scattered read 20,792 302 15 0.43 User I/O
-------------------------------------------------------------------------------
Chain 1:
-------------------------------------------------------------------------------
Oracle session identified by:
{
instance: 1 (nbdb.nbdb)
os id: 29735
process id: 25, oracle@TMC220
session id: 1202
session serial #: 40881
}
is not in a wait:
{
last wait: 0.023747 sec ago
blocking: 0 sessions
short stack: ksedsts()+480
grp()+1344
()+624
-opiino()+1680
wait history:
1. event: 'kfk: async disk IO'
time waited: 0.000015 sec
wait id: 51510 p1: 'count'=0x1
p2: 'intr'=0x0
p3: 'timeout'=0xffffffff
* time between wait #1 and #2: 0.067130 sec
2. event: 'kfk: async disk IO'
time waited: 0.000011 sec
wait id: 51509 p1: 'count'=0x1
p2: 'intr'=0x0
p3: 'timeout'=0xffffffff
* time between wait #2 and #3: 0.064147 sec
3. event: 'kfk: async disk IO'
time waited: 0.000016 sec
wait id: 51508 p1: 'count'=0x1
p2: 'intr'=0x0
p3: 'timeout'=0xffffffff
}
kfk: async disk IO==》ASM
建议你 关闭recyclebin 或者 定期purge recyclebin;
HPUX平台上用ASM不太好, 还是裸设备稳定