enq: TX - index contention
作者:互联网
When running an OLTP systems, it is possible to see high TX enqueue contention on index associated with tables, which are having high concurrency from the application. This usually happens when the application performs lot of INSERTs and DELETEs concurrently. For RAC system, the concurrent INSERTs and DELETEs could happen from all the instances .
The reason for this is the index block splits while inserting a new row into the index. The transactions will have to wait for TX lock in mode 4, until the session that is doing the block splits completes the operations.
Splitter has to do the following activities:
o Allocate a new block.
o Copy a percentage of rows to the new buffer.
o Add the new buffer to the index structure and commit the operation.
In RAC environments, this could be an expensive operation, due to the global cache operations included. The impact will be more if the split is happening at a branch or root block level.
Causes:
Most probable reasons are:
o Indexes on the tables which are being accessed heavily from the application.
o Indexes on table columns which are monotonically growing. In other words, most of the index insertions occur only on the right edge of an index.
o Large data purge has been performed, followed by high concurrent insert
Identifying the Hot index:
The indexes which are having contention can be identified from the AWR reports taken during the time of the issue.
Event Waits Time(s) Avg Wait(ms) % Total Call Time Wait Class
en: TX - index contention 89,350 40,991 459 63.3 Concurrency
db file sequential read 1,458,288 12,562 9 19.4 User I/O
CPU time 5,352 8.3
Instance Activity Stats:
Statistic Total per Second per Trans
branch node splits 945 0.26 0.00
leaf node 90-10 splits 1,670 0.46 0.00
leaf node splits 35,603 9.85 0.05
And the objects can be found either from V$SEGMENT_STATISTICS or from 'Segments by Row Lock Waits' or 'Segments by ITL Waits' or 'Service ITL Waits' of the AWR reports.
Owner Tablespace Object Name Obj.Type Row Lock Waits % of Capture
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_PK INDEX 3,425 43.62
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_ST INDEX 883 11.25
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_DT INDEX 682 8.69
Segments by ITL Waits
Owner Tablespace Name Object Name Subobject Name Obj. Type ITL Waits % of Capture
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_PK INDEX 6 50.00
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_ST INDEX 3 25.00
ACSSPROD ACSS_IDX03 ACSS_ORDER_HEADER_DT INDEX 3 25.00
Solutions:
Solution here is to tune the indexes avoid heavy access on a few set of blocks.
Following are the options we could try:
- Rebuild the index as reverse key indexes or hash partition the indexes which are listed in the 'Segments by Row Lock Waits' of the AWR reports
For example:
CREATE INDEX <index name> ON <column> REVERSE;
From the Performance Tuning Guide -
Reverse key indexes are designed to eliminate index hot spots on insert applications. These indexes are excellent for insert performance. But the downside of it is that, it may affect the performance of index range scans.
http://download.oracle.com/docs/cd/B19306_01/server.102/b14211/design.htm#sthref112
The hash method can improve performance of indexes where a small number leaf blocks in the index have high contention in multiuser OLTP environment. In some OLTP applications, index insertions happen only at the right edge of the index. This could happen when the index is defined on monotonically increasing columns. In such situations right edge of the index becomes a hotspot because of contention for index pages, buffers, latches for update, and additional index maintenance activity, which results in performance degradation.
http://download.oracle.com/docs/cd/B19306_01/server.102/b14211/data_acc.htm#i2678
It's recommended to test the application performance, after rebuilding the indexes as reverse key or hash partitioned.
- Consider increasing the CACHE size of the sequences
When we use monotonically increasing sequences for populating column values, the leaf block which is having high sequence key will be changing with every insert, which makes it a hot block and potential candidate for a block split.
With CACHE SIZE (and probably with NOORDER option), each instance would use start using the sequence keys with a different range reduces the index keys getting insert same set of leaf blocks.
- Rebuild or shrink associated index after huge amount of data purge
If there is a huge amount of data purge (delete) has been done, rebuild or shrink associated index should help to reduce the wait via alter index rebuild or alter index shrink command.
- Increase PCT_FREE for the index
A higher PCT_FREE would help to avoid ITL contention on index blocks. Contention for ITL occurs when all the available ITLs within a block are currently in use and there is not enough space in the PCT_FREE area for Oracle to dynamically allocate a new ITL slot.
标签:index,contention,TX,Waits,ACSS,indexes,enq,block 来源: https://www.cnblogs.com/muzisanshi/p/12106923.html