topic Re: crash dump query in Operating System - HP-UX

crash dump query

gany59 — Mon, 11 Apr 2011 05:19:24 GMT

I checked, whether my server have the capability to save the crash dump and found out in rc.log the exit code as 0.

Configure system crash dumps
Output from "/sbin/rc1.d/S080crashconf start":
----------------------------
EXIT CODE: 0

what is the meaning of that. 0 means saving the crash dump or not.

Thanks in advance!

Re: crash dump query

Raj D. — Mon, 11 Apr 2011 05:51:51 GMT

Gany,
check in /etc/rc.config.d/savecrash
if SAVECRASH value set to 1, then it will save the crash dump. Seems it is currenly not configured to capture crash dump,

SAVECRASH = 1
SAVECRASH_DIR = /var/adm/crash #(optional)

- Also check output of # crashconf -v

Cheers,
Raj.

Re: crash dump query

Dennis Handly — Mon, 11 Apr 2011 07:09:56 GMT

>what is the meaning of that? 0 means saving the crash dump or not.

0 means the script was successful. Whether this is connected to saving crash dumps requires you to look at the script. I would expect that you have to look at SAVECRASH as Raj said.

Re: crash dump query

gany59 — Mon, 11 Apr 2011 08:03:03 GMT

Hello Raj,

Thansk for ur reply, i have checked the enrty is save_crash=1. I got that picture as clear. If i said crashconf -v command, i am getting the below output, and i am unable to understand what was that. Can u pls help em out on the same.

bash-4.1# crashconf -v
CLASS PAGES INCLUDED IN DUMP DESCRIPTION
-------- ---------- ---------------- -------------------------------------
UNUSED 531781 no, by default unused pages
USERPG 49109 no, by default user process pages
BCACHE 122994 no, by default buffer cache pages
KCODE 7781 no, by default kernel code pages
USTACK 236 yes, by default user process stacks
FSDATA 0 yes, by default file system metadata
KDDATA 291991 yes, by default kernel dynamic data
KSDATA 3305 yes, by default kernel static data
SUPERPG 36047 no, by default unused kernel super pages

Total pages on system: 1043244
Total pages included in dump: 295532

Dump compressed: ON

DEVICE OFFSET(kB) SIZE (kB) LOGICAL VOL. NAME
------------ ---------- ---------- ------------ -------------------------
31:0x032000 2349940 8388604 64:0x000002 /dev/vg00/lvol2
----------
8388604

Re: crash dump query

Ismail Azad — Mon, 11 Apr 2011 08:31:01 GMT

Hi gany,

These are classes of memory you see on the crashconf output. Remember, there are files for crashdumps depending on when the system panics and when the crash is preserved or in lucid terms , I would just say on the way down and on the way up.

You can do a selective dump by choosing classes of kernel memory and this is done by using the -i option on crashconf and to exclude a class of memory, you would use -e. These options are typically used in the context of what is called as a "selective dump".

Regards
Ismail Azad

Re: crash dump query

Ismail Azad — Mon, 11 Apr 2011 09:30:17 GMT

Hi gany,

Ok.... There are different kinds of dump configurations and on modern integrity servers you have live dumps and concurrent dumps as well. When a dump is configured you can select one or more classes of memory on crashconf to dump. This is called as a selective dump and if you want to dump accross multiple devices you can go for a concurrent dump. A live dump can also be done without simulating a crash.

These are the steps....

lvcreate -s y -C y -r n -L -n dump vg00
crashconf -v /dev/vg00/dump

When you execute these commands crashconf -v will also now reflect /dev/vg00/dump. Normally the amount of space you want to dump should be there at the minimum unless you are using a compressed dump.

A compressed dump is dependent on a parameter called as dump_compress_on and concurrent dump on dump_concurrent_on which can be checked with the command kctune.

kctune -v dump_compress_on
kctune -v dump_concurrent_on

So first you can decide what kind of dump to use whether selective, compressed, concurrent or on an integrity server a livedump.

What you see configured in your output is only the primary swap which is taken by default. After the steps I have shown you , you should also see /dev/vg00/dump in your crashconf output.

You also need_to_check /etc/rc.config.d/crashconf and verify variables there as well. That is the up and down I was talking about as savecrash is for the preservation.

Regards
Ismail azad

Re: crash dump query

Raj D. — Mon, 11 Apr 2011 12:00:46 GMT

Gany,
Seems to be you have 8GB default dump device enabled with savecrash 1, if any system crash happen it will store the crash data in /var/adm/crash/crash.x/ directory.

/dev/vg00/lvol2 is the current dump device.

Total pages in the system showing around 4GB (taking 4k memory page size) in (crashconf -v output) , so you have enough space to keep the crash data in case of a crash happens.

More about crash dump , pls check out:
http://docs.hp.com/en/951/syscrash.html

Hth,
Raj.