Operating System - HP-UX
1748123 Members
3238 Online
108758 Solutions
New Discussion юеВ

Re: hp-ux 11.31 dump device using San external LUN

 
SOLVED
Go to solution
jose da silva
Frequent Advisor

hp-ux 11.31 dump device using San external LUN

Hello

I have successfully created secondary swap space on an external SAN LUN using the persistant devices. Now having several issues to configure an external lun for DUMP. The BL890c has 400GB of ram so /var/adm/crash is not large enough to hold a dump for analysis. Is it possible to redirect the dump " crash " to a different location ? What are the best practices?
18 REPLIES 18
Chandrahasa s
Valued Contributor

Re: hp-ux 11.31 dump device using San external LUN

Hi,

I am facing server problem....

What are those several problems??

You can specify crash path on file

/etc/rc.config.d/savecrash

You need to specify path like

SAVECRASH_DIR=/var/adm/crash

Your memory size is 400GB so you also need to consider dump speed.
HPux v3 came up with some new feature which will help to reduce dump time.

chandra

Assigning points will motivate each one in this forum
jose da silva
Frequent Advisor

Re: hp-ux 11.31 dump device using San external LUN

Hello Chandra

Thanks for your quick responce. I think you have given me the correct hint. It seems I will have to assign /var/adm/crash path to a different location for example /dump. What I would like to know if I can assign a link between /var/adm/crash to /dump or should I specify the new dump path /dump in /etc/rc.config.d/crashconf ?

Warm Regards,
Jose

Re: hp-ux 11.31 dump device using San external LUN

Seems to be some confusion around terminology here... /var/adm/crash is _not_ the dump device... it is a filesystems where dumps are written out to on reboot *following a dump* _from_ the dump device.

The dump device is a raw partition that HPUX can write a dump to without having to worry about complex kernel functions like writing to filesystems etc... it is never a filesystem.

By default on most HPUX systems, the dump device is primary swap... you can check this using:

lvlnboot -v

HTH

Duncan

I am an HPE Employee
Accept or Kudo
likid0
Honored Contributor

Re: hp-ux 11.31 dump device using San external LUN

You can also check your crash devices using the crashconf command:

crashconf
Crash dump configuration has been changed since boot.

CLASS PAGES INCLUDED IN DUMP DESCRIPTION
-------- ---------- ---------------- -------------------------------------
UNUSED 1919494 no, by default unused pages
USERPG 1051893 no, by default user process pages
BCACHE 409946 no, by default buffer cache pages
KCODE 7140 no, by default kernel code pages
USTACK 8640 yes, by default user process stacks
FSDATA 0 yes, by default file system metadata
KDDATA 740852 yes, by default kernel dynamic data
KSDATA 3098 yes, by default kernel static data
SUPERPG 35735 no, by default unused kernel super pages

Total pages on system: 4176798
Total pages included in dump: 752590

Dump compressed: ON

DEVICE OFFSET(kB) SIZE (kB) LOGICAL VOL. NAME
------------ ---------- ---------- ------------ -------------------------
31:0x006000 1547124 4194300 64:0x000002 /dev/vg00/lvol2
31:0x006000 5741428 4194300 64:0x000010 /dev/vg00/lvswap2
----------
8388600


It's better not to use a symbolic link, you should change the dir in the /etc/rc.config.d/crashconf file
Windows?, no thanks
jose da silva
Frequent Advisor

Re: hp-ux 11.31 dump device using San external LUN

Hello Colleagues

Duncan you are correct I confused the issue between /var/adm/crash and dump space. I am going to give full details on the problem I am having which I have found a workaround.

The HP-UX 11.31 blade has a total of 400 GB of memory so I created in vg00 swap space of 50GB "vol2" and decided to do the following:-
1. Create secondary swap space of 800 GB on an external SAN LUN " HiTachi"
2. Create a dump space of 400GB on an external SAN LUN.

I think the problem is that the Storage Administrator configured disk sizes of 70GB each,that is for swap "70GBx12disks=840GB" and for Dump "70GBx6disk=420GB". We created VG's with -p 32 -s 65535
So when we create the volume groups and try to create the logical volumes with Strict and Contiguous allocation policies this process fails " Not emough free physical extents available. Failure possibly caused by Contiguous allocation policy"
When I tested using only a 70GB lun with a size of 70GB disk " that is one disk only" the process passes no problems. So due to this fact it seems that for swap and dump it must be one LUN size consisting of one disk " NO EXTENDING of VG or LV's"

Please can you verify and give me advise on this problem. Perhapps I am missing something in HP-UX 11.31 version. Also since primary swap lvol2 was created with LV version 1 should we try to use LV version 2 and above?

Warm Regards,
Jose da SIlva
likid0
Honored Contributor

Re: hp-ux 11.31 dump device using San external LUN

Hi,

You don't need 800 gigs of swap space, at most if you didn't want to use pseudo swap, you could create a 400 gig swap device(it's allways better to have several devices on different controllers/arrays with the same priority, than one big chunk).

If you use pseudo swap(default behaviour), you would be ok with much less space
Windows?, no thanks
likid0
Honored Contributor

Re: hp-ux 11.31 dump device using San external LUN

Pseudo-Swap

Pseudo-swap space allows for the use of system memory as a third type of swap space. That is, HP-UX swap space can also consist of up to seven-eighths (87.5%) of system memory capacity.

For example, a computer with one GB of system memory and one GB of device and file system swap, can run up to 1.87 GB of processes. If any process attempts to grow or be created beyond this extended threshold, the process will fail.

When using pseudo-swap, since more processes can be created, the system load increases, causing more paging and deactivation activity.

By default, pseudo-swap space is configured to be available. If you do not wish to make use of it, you will need to re-set the tunable system parameter, swapmem_on, to 0 ("off"). (To modify a configurable parameter, see "Making Adjustments to Your System" in Chapter 1.)
Windows?, no thanks
chris huys_4
Honored Contributor

Re: hp-ux 11.31 dump device using San external LUN

Hi Jose,

Its not necessary to have a 400GB /var to "hold" a 400GB "full" crashdump.

The crashdump gets, when COMPRESS=1 is set in /etc/rc.config.d/savecrash, gzip'd when written from "physical dump device(s)" to "/var/adm/crash".

Therefor the /var filesystem, would need to have 80GB, 20% of full crashdump, of free diskspace available to save a "full" crashdump in /var/adm/crash.

Increase /var to f.e. 88GB, 80GB for dump and 8 GB for the rest of /var, and do a test with a full crashdump, i.e. change first with crashconf the parameters so that everything gets dumped, #crashconf -i all;crashconf -v, and then toc the system from the mp. MP>CM>tc.

NOTE: Default, when a system "crashes", only "selective" crashdump are dumped, which would only dump a fraction, check crashconf -v output, of the 400GB of RAM to physical dumpdevice(s), however in some cases, HP support will ask for a full crashdump and for this occasion, the system should indeed be "prepared" upfront.

For the "contiguous problem" of the question. Yes, I think, you cant have the dump logical volume, "cross" physical disk/lun boundaries, but on the other hand, you can for each of youre 6 "70GB" "dump" "luns", create a seperate "dump" "vol", i.e. dumpvol1 of 70Gb/dumpvol2 of 70G.. till dumpvol6 of 70Gb, and add all this "dumpvolx" to /etc/fstab, f.e. /dev/vg00/dump1 / dump defaults 0 0, and also to crashconf, with crashconf -s , also delete primary dumpvol, /dev/vg00/lvol2 first from the list of dumpvols defined in crashconf, crashconf -ds /dev/vg00/lvol2. If then a crashdump occurs, the dump should be written consecutively(sp?) to each specific dumpdevice that it encounters in crashconf.

NOTE: In HP-UX 11.31, it might even be that the dumpvolumes dont need to part anymore of /dev/vg00, well with vxvm this is the case, didnt test it with LVM. The dumpluns, where the dumpvolumes are created upon, however should have been made "lif'able" "upfront", through f.e. pvcreate -B /dev/rdisk/disk114;mkboot -e -l /dev/rdisk/disk114, to have them be part of the "persistent dump device list" (crashconf -v) , which is a absolute necessity, to have the dumpconfiguration "survive" reboots.

Greetz,
Chris
Dennis Handly
Acclaimed Contributor

Re: hp-ux 11.31 dump device using San external LUN

>Daniel: If you use pseudo swap (default behaviour), you would be ok with much less space

The default and only behavior for 11.31.