Increasing DUMP space: HP-UX 11.0

[Kevin Urschatz]

I have been conflicting opinions on the following scenario:

I have an internal 18.6 Gbyte disk in an L-Class using HP-UX 11.0.

Currently I have 2 Gbyte of Swap/Dump and have recently upgrade to 3 Gbytes of Memory. Thus I want to increase my Dump space to 3.5 Gbytes to accomodate a full dump.


I've been told that I can simply define additional dump space using some free space on the drive.

But then I've been advised that yes I can add dump space but it has to be at the beginning of a drive. So either I add a 2nd drive and define the additional dump space there or I rebuild the configuration of my current drive.

Any comments?

[Victor BERRIDGE]

I dont know of beginning of a drive... since Ive crated some on external RAID cabs...
But sure the space has to be contiguous...

[Kevin Urschatz]

Don't bother replying.

I found another message in the forum that answered my question.

Thank you

[Stefan Farrelly]

But under HP-UX 11 you dont need dump space the size of memory as an HP-UX crashdump will only be a subset of memory - ie, its a lot more intelligent than the old HP-UX 10 dumps, your dump should only be a smallish % of RAM size so no need to increase your dump space.

[Rita C Workman]

The dump you are referring to is lvol2. And it is my understanding that you would have to re-ignite the box to increase lvol2. This is set at the time of system installation. Since this dump area requires contiguous space, it would be virtually impossible to increase it after the box has been set up and ensure contiguous space.

[James R. Ferguson]

Kevin:

...and here's an interesting document on dumpspace and SuperDome servers. In part:

For a full dump, simply configure dump space (raw and file system) equal to the amount of memory, plus about 0.03 percent extra (minimum 32k) for overhead.

For a selective dump, there is no real cut-and-dried formula. Space needed depends on what kind of load is running on the system. A general guideline is to configure about 25 percent of memory. This is sufficient for most systems. The way to really determine memory required for a selective dump is to run the system with its normal load, then use the crashconf(1M) utility to see how many pages will be included in a dump.

...JRF...

[Kevin Urschatz]

True that under 11.0 that a selective dump is done thus requiring less space for a dump.

However, there are still instances where a full dump is required. So you need a dump area at least as large as RAM plus about 20% for the headers in the dump.

After a bit more reading it's now my understanding that I can specify in /etc/rc.config.d/savecrash a mount point to place a dump.

My question now is can I simply add an lvol of adequate size, place it's name in the savecrash file and then reboot the system force it to read the savecrash file? At this point will it then place the dump in that lvol? How can I confirm? (will crashconf -v display it as dump area)

Also, apparently we can create an lvol and the define it in /etc/fstab as dump. When yoiu run crashconf -a it's supposed to read /etc/fstab and configure any lvol identified as dump. Tried it and then ran crashconf -v to see if it recognizes it. No luck.

Has anyone done this?
Can you provide a sample of an entry in /etc/fstab for a dump device.

Did crashconf -v show the new lvol as a dump device? If not, what did you do to confirm that the new dump area was valid?

Thanks for all the suggestions so far,
Kevin

[CHRIS_ANORUO]

Yes Kevin it has been done before.
Do the following:
1)Install the disk
2) Create the logical volume that will be used as the dump volume. This volume should be contiguous: lvcreate -n dump -L 64 -C y /dev/vgNN
3)Specify that the logical volume, dump, will be used as the dump volume: lvlnboot -d /dev/vgNN/dump.

[Kevin Urschatz]

Chris,

Can you do this for multiple dump volumes?
Do the volumes have to be in the first 2Gbytes of the disk?