操作系统诊断工具truss, pstack, and pmap truss Truss is an UNIX utility to trace Unix System Calls in Solaris platform. Truss utility is very useful to understand complex problems at OS level. As Truss utility generates enormous amount of data, Oracle Database Instrumentation is always a first step to troubleshoot a problem. If the problem cannot be distilled by Oracle Database instrumentation, then the use of OS tools such as truss, pstack etc are required.
To trace a process and print minimal information: truss –p <pid> Example: truss –p 23898 To trace a process, follow its children and print minimal information: truss –f –p <pid> Example: truss –f –p 23898 To trace a process, print timestamp and print minimal information: truss –d –p <pid> Example: truss –d –p 23898 To trace a process, send output to a file and print minimal information: truss –o /tmp/truss.out –p <pid> Example: truss –o /tmp/truss.out –d –p 23898
Utility in UNIX platform Platform Name Solaris truss HP-UX TUSC( truss) Linux strace AIX truss
Truss of DBWR with –d –D flag $ truss -d -D -p 1473 |more Truss of DBWR with –d –D flag $ truss -d -E -p 19001 |more Truss of DB startup with –f flag $ truss -d -E –f -p 2522 |more 在Linux下面 $strace -ttT -p 5164
AIX & HP HP-UX operating system provides TUSC a tool similar to truss utility. But, truss is soft linked to tusc utility and supports many options as in truss utility in Solaris. Essentially, discussions for Solaris, applies to HP-UX platform too. In AIX, truss is available too. But, truss in AIX does not support -E flag. There is no easy way to measure amount of time elapsed in the system call itself using truss utility(at least, not that I know of). pmap pmap command can be used to understand virtual memory mapping of an UNIX process, memory usage, and attributes of memory area etc. In AIX, this tool is named as procmap In Linux, & HP-UX, pmap is available pmap –xs output $ pmap -x 2540 |more
pmap #! /bin/ksh pid=$1 (( cnt=1000 )) while [[ $cnt -gt 0 ]]; do date pmap -x $pid pstack $pid echo $cnt (( cnt=cnt-1 )) sleep 10 done ---A small shell script, to dump Memory map and stack of a process, in a loop,every 10 seconds
pstack Utility pstack can print the current execution stack of a process. Oracle Database processes are instrumented to measure performance and most performance issues can be resolved using that instrumentation. Database processes are usually in one of these three states, either executing a piece of code, waiting for an event such as I/O, or waiting in CPU scheduling queue to be scheduled. To measure performance of a program, optimal task is to alter the session to enable sql trace and execute the program. But, that is not always possible in a production environment. So, if a program is running longer in production, you can review the ASH data to measure the waits. But, if the program is not suffering from waits and spending time executing in CPU, how do you measure the performance of a program? Utility pstack comes handy in these situations. Essentially, current execution stack of a process can be printed using pstack utility. Using pstack in a loop, you can generate many samples of execution stack of a process; with some aggregation, you can understand the performance of a process better. In AIX, this tool is named as procstack pstack output $pstack 2544
Pfiles: ? pfiles can be used to associate this file ids with file names. ? Pfiles lists the files currently opened by a process. In few unix platform, this can be achieved by lsof command also.
Oradebug utility can be used to print the call stack too. In Listing 1-14, call stack of the process is printed in the current process, using setmypid command. Listing 1-14 oradebug SQL> oradebug setmypid Statement processed. SQL> oradebug short_stack ksedsts()+1123<-ksdxfstk()+33<-ksdxen_int()+5127<- ksdxen()+14<-opiodr()+1075<-ttcpip()+1433<-opitsk()+1536<- opiino()+1653<-opiodr()+1075<-opidrv()+814<-sou2o()+87<- opimai_real()+537<-ssthrdmain()+334<-main()+203<- _start()+108 SQL> oradebug short_stack qksedsts()+1123<-ksdxfstk()+33<-ksdxen_int()+5127<- ksdxen()+14<-opiodr()+1075<-ttcpip()+1433<-opitsk()+1536<- opiino()+1653<-opiodr()+1075<-opidrv()+814<-sou2o()+87<- opimai_real()+537<-ssthrdmain()+334<-main()+203<- _start()+108