One month ago a colleague asked me a question: why the "file-size" in output of "sar -v" is different with "lsof | wc -l"? For example: # sar -v 09:23:31  proc-sz     inod-sz     file-sz     thrd-sz    09:23:32  82/262144   0/201       933/11215   321/524288 # lsof | wc -l      767 From above output we can find that lsof reports a smaller number than file-sz (767 vs 933). He thought the file-size means the number of open files in the system, and the lsof also reports files opened by processes in the system. So it... [More]

Tags:  sar lsof kdb aix

  GCC and bash depend on libintl.a.   We may download gettext from http://www.bullfreeware.com/, and run: rpm -ivh gettext-0.17-7.aix6.1.ppc --nodeps   then we may find   libintl.a in  /opt/freeware/lib.

Tags:  aix libintl gcc

This is the link for my article: http://www.ibm.com/developerworks/aix/library/au-aix-jfs2-inode/index.html http://www.ibm.com/developerworks/aix/library/au-aix-j2inode-CacheSize/index.html

For large heavy-used jfs2 file system, it will take long time to umount it. But if you manually run "sync" before umount, you will find the total time of "sync+umount" is much less than umount merely.   But theoretically both syncd and umount itself will do same work as sync..... Why  does sync has better performance?   The reason is that sync will use multi-threads when it update "dirty" data of jfs2 file system. But syncd just use one thread for one gfs type.   Such design is reasonable. When... [More]

Tags:  umount syncd jfs2 sync

  This tool will show detail info about a jfs2 fs. You can run it regardless of whether the fs is mounted.   For example: ./j2info /dev/fslv00 > fs.log   Then it will generate 3 files: fs.log, inoblk.log, blks.log.   In fs.log we have metadata and detail info for all inodes.   From inoblk.log we can know how many blocks each inode has used:   #more inoblk.log  ino     blocks 2       0 3       0 4       1 32      0 64      0... [More]

#svmon -G -O unit=auto     Unit: auto --------------------------------------------------------------------------------------                         size       inuse             free         pin               virtual          available   mmode memory        1.00G     971.20M       52.8M    ... [More]

https://www.ibm.com/developerworks/community/blogs/aixpert/entry/installing_debain_7_and_fedora_18_on_power7?lang=en A good article.  

I find many people are confused with this question. They usually think it is " maxuproc". In fact " maxuproc" means maximum number of processes per user. And the  maximum number of processes for aix 5.3/6.1/7.1 is 256k.

  For example: #oslevel -s 6100-06-05-1115 ./blvinfo -o /dev/hd5           bi_start     = 0x3600         bi_size      = 0xf90000         bd_size      = 0xf70000         ram FS start = 0x8c1140         ram FS size  = 0x6aed6b         sba_start    = 0xf73600         sba_size     = 0x20000         sbd_size  ... [More]

We all know that dump reason code 300 means Data Storage Interrupt, 400 means Instruction Storage Interrupt, and 700 means Program Interrupt. But why number 300/400/700 are selected instead of other numbers?   In fact these numbers are not selected randomly. They are the interrupt vector of DSI/ISI/PI.   (0)> dc 300 80 00000300     mtsprg    SPRG2,stkp 00000304     mfsprg    stkp,SPRG0 00000308         ld    stkp,368(stkp) 0000030C    ... [More]

We can use skeyctl to view and change settings related to storage key.   Usage: skeyctl [-D]         skeyctl [-h <nhkeys>] [-u <nukeys>/off] [-k on/off/default] [-x kkey ]        skeyctl [-v [now|default|boot]                where:                        -h <nhkeys>   # number of hardware keys (0, 2-31)                    ... [More]

http://www.ibm.com/developerworks/aix/tutorials/au-interceptcall/index.html interceptcall.pdf   In this article I have showed how to get individual system call number using dbx. We can also get all system calls  from procfs as follows: #include <stdio.h> #include <stdlib.h> #include <fcntl.h> #include <sys/procfs.h>   #define BUFSIZE 1024*64 int  main() {         prsysent_t *prsysp;         char buffer[BUFSIZE];         int fd,... [More]

  Sometimes we need to extract odm in boot lv and examine them. So I add a feature for blvinfo to automatically do such work, like:   # ./blvinfo -m /dev/hd5         bi_start     = 0x3600         bi_size      = 0x1920000         bd_size      = 0x1900000         ram FS start = 0x95c4f0         ram FS size  = 0xefe100         sba_start    = 0x1903600  ... [More]