pstack, truss etc to understand deeper issues in Oracle database

pstack, truss, etc
By
Riyaj Shamsudeen

©OraInternals Riyaj Shamsudeen 2
Me
 19 years using Oracle products/DBA
 OakTable member
 Oracle ACE Director
 Certified DBA versions 7.0,7.3,8,8i,9i
&10g
 Specializes in RAC, performance tuning,
Internals and E-business suite
 Chief DBA with OraInternals
 Co-author few books
 Email: rshamsud@orainternals.com
 Blog : orainternals.wordpress.com
 Web: www.orainternals.com

Agenda
 Truss with few examples.
 Demos of truss (hopefully)
 Proc tools intro: pmap, pfiles etc.
 pstack

Truss
 Truss is to trace UNIX system calls.
 Use only if Oracle instrumentation is not sufficient
to debug.
 Nearly, all platforms provides tools similar to Truss
utility in Solaris.

Truss - trace system calls and signals
truss –p 28393
…
fstat(18, 0xFFBFA100) = 0
lseek(17, 0x0216B000, SEEK_SET) = 0x0216B000
write(17, "C8021686 )00 )D0000".., 4096) = 4096
lseek(17, 0x0219D000, SEEK_SET) = 0x0219D000
read(17, "00000101001FF 00".., 4096) = 4096

Truss
Description:
The truss utility traces the system calls and the signal process receives.
Options:
truss [-fcaeildD] [ - [tTvx] [!] syscall ,...] [ - [sS] [!]
signal ,...] [ - [mM] [!] fault ,...] [ - [rw] [!] fd ,...] [ -
[uU] [!] lib ,... : [:] [!] func ,...] [-o outfile] command |
-p pid...
Solaris – truss
Hpux- tusc
Linux – strace
AIX - truss

Truss
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

Truss – Word of caution
At every system call, truss inspects the process. This
*potentially* could slow down the process.
So, Truss critical processes, only when it is necessary to
do so.

Truss – Few outputs
$ truss -d -E -p 1873
Base time stamp: 1310009834.7781 [ Wed Jul 6 22:37:14 CDT 2011 ]
0.0124 0.0000 semtimedop(7, 0xFFFFFD7FFFDF9328, 1, 0xFFFFFD7FFFDF9340) Err#11 EAGAIN
0.1130 0.0000 mmap(0xFFFFFD7FFC1DE000, 65536, PROT_READ|PROT_WRITE, MAP_PRIVATE|
MAP_FIXED, 7, 0) = 0xFFFFFD7FFC1DE000
Time stamp displacement
From base timestamp.
Seconds.fraction of sec
Elapsed time in the system call.
Semtimedop is the call used by the database processes to sleep
Waiting for an event or timer expiry.

PMON behavior
…
0.7998 0.0002 open("/proc/1456/psinfo", O_RDONLY) = 40
0.7998 0.0000 read(40, "0000201000B00500".., 416) = 416
0.7999 0.0001 close(40) = 0
0.7999 0.0000 read(40, "0000201000B20500".., 416) = 416
0.8000 0.0001 close(40) = 0
0.8001 0.0001 read(40, "0000202000B60500".., 416) = 416
0.8001 0.0000 close(40) = 0
…
PMON is checking the status of database connection processes.
This is how PMON identifies dead process to cleanup.

PMON – detect killed process
…
20.8818 0.0000 open("/proc/2372/psinfo", O_RDONLY) Err#2 ENOENT
20.8826 0.0008 pollsys(0x0EA992B0, 3, 0xFFFFFD7FFFDFAFD0, 0x00000000) = 0
20.8828 0.0002 write(42, "04 Z00060000000".., 1114) = 1114
20.8838 0.0010 write(37, "04 Z00060000000".., 1114) = 1114
20.8843 0.0005 pollsys(0x0EA992B0, 3, 0xFFFFFD7FFFDFB0E0, 0x00000000) = 2
Process 2372 was killed and ENOENT thrown. PMON
detected the process death and cleaned up the resources.

DBWR truss
…
$ truss -d -D -p 1473 |more
Base time stamp: 1327357172.8340 [ Mon Jan 23 16:19:32 CST 2012 ]
…
/1:semtimedop(7, 0xFFFFFD7FFFDFD188, 1, 0xFFFFFD7FFFDFD1A0) (sleeping...)
/1:3.2222 3.0003 semtimedop(7, 0xFFFFFD7FFFDFD188, 1, 0xFFFFFD7FFFDFD1A0) Err#11 EAGAIN
…
…
…
Semtimedop call is used by DBWR to sleep for 3 seconds waiting on
semaphore. Wakesup at semaphore activity or timeout.

Foreground reads
…
13.2066 0.0001 pread(256, "06A20098 AC001A9 . 50".., 8192, 0x14030000) = 8192
13.3202 0.0001 pread(261, "06A200C6 ?C001 P . 50".., 8192, 0x1418C000) = 8192
13.3215 0.0000 pread(262, "06A200 3 EC00101 . 50".., 8192, 0x14666000) = 8192
13.3250 0.0000 pread(257, "06A20099 >C001C6 * 50".., 8192, 0x14532000) = 8192
13.3618 0.0001 pread(257, "06A200CE FC00101 . 50".., 8192, 0x1479C000) = 8192
13.4259 0.0001 pread(263, "06A200 B BC001CD . 50".., 8192, 0x16384000) = 8192
…
Foreground process reads 8K blocks from the disk in to buffer
cache. Approximately, 0.1ms per read.
Demo: trussprocess.ksh, dbf_seq_gen.sql

DB startup
ps –ef|grep oracle
truss -d -E -f -o /tmp/dbstart.lst -p 2522
4.8050 0.0000 shmget(IPC_PRIVATE, 4194304, 0660) = 58
4.8051 0.0000 shmat(58, 0, IPC_ALLOC) =
0xFFFFFD7FFC600000
4.8051 0.0000 shmctl(58, IPC_RMID, 0) = 0
4.8054 0.0000 shmget(3433015348, 536879104, 0660|IPC_CREAT|IPC_EXCL) = 59
4.8055 0.0000 shmget(3433015349, 0, 0) Err#2 ENOENT
4.8059 0.0004 shmat(59, 0x60000000, IPC_ALLOC) = 0x60000000
A sqlplus connection process was TRUSSed to get all system calls with –f flag.
Demo: startup db
A Shared memory segment is created for SGA.

Shared memory segment
ipcs –ma
Shared Memory:
T ID KEY ... CGROUP NATTCH SEGSZ
m 59 0xcc9fa834... Oinstall 54 536879104
M 1 0xea42bf0c... oinstall 37 285220864
Shared memory segment was created with that shmget call, 59 shm id.
Demo: startup db
0xcc9fa834 = 3433015348
4.8054 0.0000 shmget(3433015348, 536879104, 0660|IPC_CREAT|IPC_EXCL) = 59

PMON startup
2522 is sqlplus connection process, a
double fork to pmon pid 2540
2522: 6.4786 0.0024 fork1() = 2538
2538: 6.4786 0.0000 fork1() (returning as child ...)= 2522
2538: 6.4812 0.0000 getpid() = 2538 [2522]
2538: 6.4819 0.0000 lwp_self() = 1
..
2538: 6.4823 0.0000 close(21) = 0
2538: 6.4824 0.0000 schedctl() = 0xFFFFFD7FFDDEE000
2538: 6.4831 0.0006 fork1() = 2540
2540: 6.4831 0.0000 fork1() (returning as child ...) = 2538
$strings /proc/2540/psinfo
Oracle
ora_pmon_solrac2

PMON attach
2522 is sqlplus connection process, a
double fork creates PMON pid 2540
2540: 6.5195 0.0000 shmget(3433015348, 0, 0) = 59
2540: 6.5195 0.0000 shmctl(59, IPC_STAT, 0xFFFFFD7FFFDFEA40) = 0
2540: 6.5196 0.0000 shmat(59, 0, 0) = 0xFFFFFD7FDCA00000
2540: 6.5197 0.0000 shmdt(0xFFFFFD7FDCA00000) = 0
2540: 6.5197 0.0000 shmget(3433015349, 0, 0) Err#2 ENOENT
2540: 6.5200 0.0002 shmat(59, 0x60000000, IPC_ALLOC) = 0x60000000
PMON process attached to the shared memory segment
through a shmat call.

Connection creation
1432 is the listener process. Double
fork creating a connection.
1432/1: 4.9712 0.0006 fork1() = 5002
5002: 4.9763 0.0000 getpid() = 5002 [1432]
…
5002: 4.9791 0.0004 fork1() = 5004
5004: 4.9820 0.0000 getpid() = 5004 [5002]
…
5004: 4.9860 0.0000 setsid() Err#1 EPERM
5004: 4.9880 0.0019 execve("/u02/app/oracle/product/11.2.0/
dbhome_2/bin/oracle", 0x004DA7A0, 0x006FDD80) argc= 2
Oracle binary executed to create oracle process.

Linux
Strace is the equivalent in Linux.
-tt is to print the timestamp with microseconds.
-T is to print the time spent in the system call.
$strace -ttT -p 5164
Process 5164 attached - interrupt to quit
09:34:35.710183 poll([{fd=27, events=POLLIN|POLLRDNORM}, {fd=29,
events=POLLIN|POLLRDNORM}], 2, 3000) = 0 (Timeout) <3.000425>
09:34:38.711308 getrusage(RUSAGE_SELF, {ru_utime={0, 10998}, ru_stime={0,
69989}, ...}) = 0 <0.000174>
09:34:38.711756 getrusage(RUSAGE_SELF, {ru_utime={0, 10998}, ru_stime={0,
69989}, ...}) = 0 <0.000174>
09:34:38.712285 open("/proc/5166/stat", O_RDONLY) = 28 <0.000154>
09:34:38.712684 read(28, "5166 (oracle) S 1 5166 5166 0 -1"..., 999) = 226
<0.000182>
09:34:38.713081 close(28) = 0 <0.000168>
09:34:38.713455 open("/proc/5170/stat", O_RDONLY) = 28 <0.000180>
09:34:38.713835 read(28, "5170 (oracle) S 1 5170 5170 0 -1"..., 999) = 225
<0.000176>
09:34:38.714216 close(28) = 0 <0.000174>
Read of 226 bytes took 1.82 ms at 09:34:38.712684

AIX & HP
 Truss is available in AIX & HP too.
 In HP-UX, truss is called tusc; Usually, there are
soft links to tusc named truss.
 man pages on truss or tusc should give all options
available.
In AIX, -E flag, which is very useful is not available.

truss & pfiles
Truss:
write(18, " 9 8 7 1 0 o b jn <".., 21) = 21
fstat(18, 0xFFBFA058) = 0
write(18, " 9 8 7 2 0 R > >n s".., 18) = 18
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.

pfiles
Using these device numbers and
Inode numbers, file names can be mapped.
pfiles 28393
28393: ar60runb P_CONC_REQUEST_ID=2452107 STARTDATE='012006'
ENDDATE='122006'
Current rlimit: 4096 file descriptors
0: S_IFIFO mode:0000 dev:272,0 ino:7325504 uid:11175 gid:100 size:0
...
17: S_IFREG mode:0644 dev:233,63004 ino:895242 uid:11175 gid:100 size:
102522880
O_RDWR|O_CREAT|O_TRUNC
18: S_IFREG mode:0644 dev:233,63004 ino:895305 uid:11175 gid:100 size:
25491841
O_RDWR|O_CREAT|O_TRUNC
This is the file_id
In the truss output
This is the device id
Of the form minor,major
Inode number

Pfiles & proc tools
Many tools available, aka proc tools
pflags, pcred, pldd, psig, pstack, pfiles, pwdx,
pstop, prun, pwait, ptree, ptime
WARNINGS
The following proc tools stop their target processes while inspecting them
and reporting the results: pfiles, pldd, pmap, and pstack.
A process can do nothing while it is stopped. Stopping a heavily used
process in a production environment, even for a short amount of time, can
cause severe bottlenecks ..

pmap
 Process memory need to be monitored and pmap command can
give a breakdown of process memory.
 Useful in Oracle environments to differentiate SGA and PGA of
a connection process.
 Some information is cryptic though.
In AIX, this tool is named as procmap
In Linux, & HP-UX, pmap is available

pmap <pid>
$ pmap -x 2540 |more
2540: ora_pmon_solrac2
Pmap prints a Nice memory map
of the Process. Various heaps and
Stacks are printed here
Address Kbytes RSS Anon Locked Mode Mapped File
0000000000400000 232736 37864 - - r-x-- oracle
000000000E757000 1424 476 144 - rw--- oracle
000000000E8BB000 156 32 32 - rw--- oracle
000000000E8E2000 1972 1132 1124 - rw--- [ heap ]
0000000060000000 526336 321508 - - rwxs- [ dism
shmid=0x3b ]
FFFFFD7FFCAA0000 64 - - - rwx-- [ anon ]
FFFFFD7FFCABE000 72 8 8 - rw--- [ anon ]
FFFFFD7FFCAD0000 64 12 12 - rw--- [ anon ]
FFFFFD7FFCAE0000 64 20 20 - rw--- [ anon ]
..
---------------- ---------- ---------- ---------- ----------
total Kb 780044 368316 2072 -
Total memory foot print
Also printed.

pmap of pmon
Pmap –xs <pid>
...
-s flag prints Hardware Address
Translation.
Address Kbytes RSS ... Pgsz Mode Mapped File
000000000EACF000 16 - - - rw--- [ heap ]
0000000060000000 2048 2048 4K rwxs- [ dism shmid=0x7 ]
0000000060200000 378880 174080 - rwxs- [ dism shmid=0x7 ]
0000000077800000 12288 12288 - rwxs- [ dism shmid=0x7 ]
0000000078800000 12288 12288 - rwxs- [ dism shmid=0x7 ]
0000000079600000 2048 2048 rwxs- [ dism shmid=0x7 ]
000000007A200000 4096 4096 - rwxs- [ dism shmid=0x7 ]
000000007A600000 6144 6144 4K rwxs- [ dism shmid=0x7 ]
000000007AC00000 2048 2048 - rwxs- [ dism shmid=0x7 ]
000000007AE00000 2048 2048 4K rwxs- [ dism shmid=0x7 ]
…
---------------- ---------- ---------- ---------- ----------
total Kb 781064 377672
Shared memory segment that can
grow or shrink.

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.

pmap
00010000 72 72 - - r-x-- java
00030000 16 16 16 - rwx-- java
00034000 8744 8680 8680 - rwx-- [ heap ]
77980000 1224 1048 - - r--s- dev:273,2000 ino:104403
77CFA000 24 24 24 - rw--R [ anon ]
...
FF39A000 8 8 8 - rwx-- libthread.so.1
FF3A0000 8 8 - - r-x-- libc_psr.so.1
FF3B0000 184 184 - - r-x-- ld.so.1
FF3EE000 8 8 8 - rwx-- ld.so.1
FF3F0000 8 8 8 - rwx-- ld.so.1
FF3FA000 8 8 8 - rwx-- libdl.so.1
FFB80000 24 - - - ----- [ anon ]
FFBF0000 64 64 64 - rw--- [ stack ]
-------- ------- ------- ------- -------
total Kb 182352 65568 26360 -
Process initially started with
a memory usage of 182MB

pmap
00010000 72 72 - - r-x-- java
00030000 16 16 16 - rwx-- java
00034000 8808 8720 8720 - rwx-- [ heap ]
77980000 1224 1048 - - r--s- dev:273,2000 ino:104403
77CFA000 24 24 24 - rw--R [ anon ]
77F7A000 24 24 24 - rw--R [ anon ]
78000000 72 72 72 - rwx-- [ anon ]
78012000 64 64 64 - rwx-- [ anon ]
7814C000 144 144 144 - rwx-- [ anon ]
78170000 8 8 8 - rwx-- [ anon ]
78172000 8 8 8 - rwx-- [ anon ]
78174000 8 8 8 - rwx-- [ anon ]
78176000 104 104 104 - rwx-- [ anon ]
..
FF3F0000 8 8 8 - rwx-- ld.so.1
FF3FA000 8 8 8 - rwx-- libdl.so.1
FFB80000 24 - - - ----- [ anon ]
FFBF0000 64 64 64 - rw--- [ stack ]
-------- ------- ------- ------- -------
total Kb 281040 210736 171528 -

YATC (Yet Another Test Case)
Program is running for many hours. Recently there was a minor
code change to the program.
Demo: testcase2.sql

pstack $pstack 2544
2544: oraclesolrac1 (DESCRIPTION=(LOCAL=YES)
(ADDRESS=(PROTOCOL=beq)))
000000000ab1418f pevm_SUBSTR () + 12f
000000000aad49bf pfrinstr_SUBSTR () + 5f
000000000aac5880 pfrrun_no_tool () + 40
000000000aac6a6f pfrrun () + 4df
000000000ab2e3fa plsql_run () + 2ea
000000000aaa4a83 peicnt () + 143
000000000a0fba56 kkxexe () + 216
000000000447b5c7 opiexe () + 2757
0000000004d54695 kpoal8 () + ce5
0000000004472693 opiodr () + 433
0000000008e67f69 ttcpip () + 599
000000000444cfc0 opitsk () + 600
000000000445bb75 opiino () + 675
0000000004472693 opiodr () + 433
0000000004441f4e opidrv () + 32e
0000000005672197 sou2o () + 57
000000000159eac9 opimai_real () + 219
000000000568f2de ssthrdmain () + 14e
000000000159e89b main () + cb
000000000159e67c ???????? ()

pstack
In AIX, this tool is named as procstack
2544: oraclesolrac2 (DESCRIPTION=(LOCAL=YES)(ADDRESS=(PROTOCOL=beq)))
000000000ab135cc pevm_MOVC_i () + 5c
000000000aac9c9e pfrinstr_MOVC () + 2e
000000000aaa4a83 peicnt () + 143
…
00000000091a2a3b kghuwrlength () + 4b
0000000009e348f3 plsm0vc_rsz () + 43
000000000ab141ca pevm_SUBSTR () + 16a
000000000aad49bf pfrinstr_SUBSTR () + 5f
…
000000000ab26125 pevm_icd_call_common () + c5
000000000aacb5e3 pfrinstr_BCAL () + 53
000000000aaa4a83 peicnt () + 143...

Oradebug short_stack
 Oradebug short_stack also can be used to get process stack.
 Example:
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
 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>

Thank you for attending!
If you like this presentation, you will love my
2-part intensive, online RAC webinar.
http://www.orainternals.com/training/
advanced-rac-training
Week Dates Time
Week1 March 26-30 8AM – 2PM PDT
Week 2 April 9-13 8AM - 2PM PDT
©OraInternals Riyaj Shamsudeen 34

pstack, truss etc to understand deeper issues in Oracle database

More Related Content

What's hot (20)

Similar to pstack, truss etc to understand deeper issues in Oracle database (20)

More from Riyaj Shamsudeen (12)

Recently uploaded (20)

pstack, truss etc to understand deeper issues in Oracle database