数据库平台:SunOS 5.8 Generic_108528-23 sun4u sparc SUNW,Ultra-Enterprise
数据库版本:8.1.5.0.0
数据库症状:数据库响应缓慢,应用请求无法返回,业务操作陷于停顿,此时需要DBA介入并进行问题诊断及故障处理。
1. 登录数据库进行检查
首先我们登录数据库,检查故障现象。
经过检查发现,数据块的所有重做日志组除current外都处于active状态:
oracle:/oracle/oracle8>sqlplus "/ as sysdba"SQL*Plus: Release 8.1.5.0.0 - Production on Thu Jun 23 18:56:06 2005(c) Copyright 1999 Oracle Corporation. All rights reserved. Connected to: Oracle8i Enterprise Edition Release 8.1.5.0.0 - Production With the Partitioning and Java options PL/SQL Release 8.1.5.0.0 - Production SQL> select * from v$log; GROUP# THREAD# SEQUENCE# BYTES MEMBERS ARC STATUS FIRST_CHANGE# FIRST_TIM ---------- ---------- ---------- ---------- ---------- --- ---------------- ------------- --------- 1 1 520403 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 2 1 520404 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 3 1 520405 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 4 1 520406 31457280 1 NO CURRENT 1.3861E+10 23-JUN-05 5 1 520398 31457280 1 NO ACTIVE 1.3860E+10 23-JUN-05 6 1 520399 31457280 1 NO ACTIVE 1.3860E+10 23-JUN-05 7 1 520400 104857600 1 NO ACTIVE 1.3860E+10 23-JUN-05 8 1 520401 104857600 1 NO ACTIVE 1.3860E+10 23-JUN-05 9 1 520402 104857600 1 NO ACTIVE 1.3861E+10 23-JUN-059 rows selected.SQL> / GROUP# THREAD# SEQUENCE# BYTES MEMBERS ARC STATUS FIRST_CHANGE# FIRST_TIM ---------- ---------- ---------- ---------- ---------- --- ---------------- ------------- --------- 1 1 520403 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 2 1 520404 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 3 1 520405 31457280 1 NO ACTIVE 1.3861E+10 23-JUN-05 4 1 520406 31457280 1 NO CURRENT 1.3861E+10 23-JUN-05 5 1 520398 31457280 1 NO ACTIVE 1.3860E+10 23-JUN-05 6 1 520399 31457280 1 NO ACTIVE 1.3860E+10 23-JUN-05 7 1 520400 104857600 1 NO ACTIVE 1.3860E+10 23-JUN-05 8 1 520401 104857600 1 NO ACTIVE 1.3860E+10 23-JUN-05 9 1 520402 104857600 1 NO ACTIVE 1.3861E+10 23-JUN-059 rows selected. |
我们知道,当数据库发生日志切换时(Log Switch),Oracle会触发一个检查点(Checkpoint),检查点进程(Checkpoint Process,CKPT)会通知DBWR(Database?Writer)进程去执行写操作。在日志文件所保护的处于Buffer cache中的脏数据(dirty buffer)未写回磁盘之前,日志文件不能被覆盖或重用。
如果数据库异常繁忙,或者DBWR的写出过慢,就可能出现检查点未完成,Oracle却已经用完所有日志文件的情况。在这种情况下,数据库的日志无法生成,整个数据库将处于停顿状态,此时日志文件中会记录类似如下信息:
Mon Jan 23 16:11:39 2006Thread 1 cannot allocate new log,
sequence 5871Checkpoint not complete Current log# 2 seq# 5870 mem# 0:
+ORADG/danaly/onlinelog/group_2.260.600173851
Current log# 2 seq# 5870 mem# 1:
+ORADG/danaly/onlinelog/group_2.261.600173853 |
检查v$session_wait视图,我们可以从中看到很多session处于log file switch (checkpoint incomplete) 的等待。
2. 检查DBWR进程
在本案例中,所有日志组都处于active状态,那么显然DBWR的写出存在问题。
接下来让我们检查一下DBWR的繁忙程度:
SQL> ! oracle:/oracle/oracle8>ps -ef|grep ora_ oracle 2273 1 0 Mar 31 ? 57:40 ora_smon_hysms02 oracle 2266 1 0 Mar 31 ? 811:42 ora_dbw0_hysms02 oracle 2264 1 16 Mar 31 ? 16999:57 ora_pmon_hysms02 oracle 2268 1 0 Mar 31 ? 1649:07 ora_lgwr_hysms02 oracle 2279 1 0 Mar 31 ? 8:09 ora_snp1_hysms02 oracle 2281 1 0 Mar 31 ? 4:22 ora_snp2_hysms02 oracle 2285 1 0 Mar 31 ? 9:40 ora_snp4_hysms02 oracle 2271 1 0 Mar 31 ? 15:57 ora_ckpt_hysms02 oracle 2283 1 0 Mar 31 ? 5:37 ora_snp3_hysms02 oracle 2277 1 0 Mar 31 ? 5:58 ora_snp0_hysms02 oracle 2289 1 0 Mar 31 ? 0:00 ora_d000_hysms02 oracle 2287 1 0 Mar 31 ? 0:00 ora_s000_hysms02 oracle 2275 1 0 Mar 31 ? 0:04 ora_reco_hysms02 oracle 21023 21012 0 18:52:59 pts/65 0:00 grep ora_
|
DBWR的进程号是2266。
使用Top命令观察一下该进程的CPU耗用:
oracle:/oracle/oracle8>toplast pid: 21145; load averages: 3.38, 3.45, 3.67 18:53:38 725 processes: 711 sleeping, 1 running, 10 zombie, 3 on cpu CPU states: 35.2% idle, 40.1% user, 9.4% kernel, 15.4% iowait, 0.0% swap Memory: 3072M real, 286M free, 3120M swap in use, 1146M swap free PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND 11855 smspf 1 59 0 1355M 1321M cpu/0 19:32 16.52% oracle 2264 oracle 1 0 0 1358M 1316M run 283.3H 16.36% oracle 11280 oracle 1 13 0 1356M 1321M sleep 79.8H 0.77% oracle 6957 smspf 15 29 10 63M 14M sleep 107.7H 0.76% java 17393 smspf 1 30 0 1356M 1322M cpu/1 833:05 0.58% oracle 29299 smspf 5 58 0 8688K 5088K sleep 18.5H 0.38% fee_ftp_get 21043 oracle 1 43 0 3264K 2056K cpu/9 0:01 0.31% top 20919 smspf 17 29 10 63M 17M sleep 247:02 0.29% java 25124 smspf 1 58 0 16M 4688K sleep 0:35 0.25% smif_status_rec 8086 smspf 5 23 0 21M 13M sleep 41.1H 0.24% fee_file_in 16009 root 1 35 0 4920K 3160K sleep 0:03 0.21% sshd2 25126 smspf 1 58 0 1355M 1321M sleep 0:26 0.20% oracle 2266 oracle 1 60 0 1357M 1317M sleep 811:42 0.18% oracle 11628 smspf 7 59 0 3440K 2088K sleep 0:39 0.16% sgip_client_ltz 26257 smspf 82 59 0 447M 178M sleep 533:04 0.15% java |
我们注意到,2266号进程消耗的CPU不过0.18%,显然并不繁忙,DBWR并不繁忙,但是检查点无法完成,那么我们可以判断,瓶颈就很可能出现在IO上。
3. 检查IO状况
我们可以使用IOSTAT工具检查系统IO状况:
gqgai:/home/gqgai>iostat -xn 3 extended device statistics r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device ...... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t6d0 1.8 38.4 32.4 281.0 0.0 0.7 0.0 16.4 0 29 c0t10d0 1.8 38.4 32.4 281.0 0.0 0.5 0.0 13.5 0 27 c0t11d0 24.8 61.3 1432.4 880.1 0.0 0.5 0.0 5.4 0 26 c1t1d0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 9.1 0 0 hurraysms02:vold(pid238) extended device statistics r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device ........ 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t6d0 0.3 8.3 0.3 47.0 0.0 0.1 0.0 9.2 0 8 c0t10d0 0.0 8.3 0.0 47.0 0.0 0.1 0.0 8.0 0 7 c0t11d0 11.7 65.3 197.2 522.2 0.0 1.6 0.0 20.5 0 100 c1t1d0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 hurraysms02:vold(pid238) extended device statistics r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device ........ 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t6d0 0.3 13.7 2.7 68.2 0.0 0.2 0.0 10.9 0 12 c0t10d0 0.0 13.7 0.0 68.2 0.0 0.1 0.0 9.6 0 11 c0t11d0 11.3 65.3 90.7 522.7 0.0 1.5 0.0 19.5 0 99 c1t1d0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 hurraysms02:vold(pid238) extended device statistics r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device ........ 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t6d0 0.0 8.0 0.0 42.7 0.0 0.1 0.0 9.3 0 7 c0t10d0 0.0 8.0 0.0 42.7 0.0 0.1 0.0 9.1 0 7 c0t11d0 11.0 65.7 978.7 525.3 0.0 1.4 0.0 17.7 0 99 c1t1d0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 hurraysms02:vold(pid238) extended device statistics r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device ........ 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t6d0 0.3 87.7 2.7 433.7 0.0 2.2 0.0 24.9 0 90 c0t10d0 0.0 88.3 0.0 436.5 0.0 1.8 0.0 19.9 0 81 c0t11d0 89.0 54.0 725.4 432.0 0.0 2.1 0.0 14.8 0 100 c1t1d0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 hurraysms02:vold(pid238) |
在以上输出中,我们注意到,存放数据库的主要卷c1t1d0的繁忙程度始终处于99~100,而写速度却只有500K/s左右,这个速度是极为缓慢的。
根据IOSTAT的手册:
(%b percent of time the disk is busy (transactions in progress) Kw/s kilobytes written per second) |
根据我们的常识,T3盘阵通常按Char写速度可以达到10M/s左右,以前测试过一些Tpcc指标,可以参考:www.eygle.com/unix/Use.Bonnie.To.Test.IO.speed.htm。
而正常情况下的数据库随机写通常都在1~2M左右,显然此时的磁盘已经处于不正常状态,经过确认的确是硬盘发生了损坏,Raid5的Group中损坏了一块硬盘。
经过更换以后系统逐渐恢复正常。