filesystem size for /crash

[Global Server Operation]

I'm setting configuring a new server which is a rp5470 running 11i. 32 bit.It also contains 4096MB of memory. Should I set up a seperate filesystem for /crash and what size should it be set at?

All reply are appreciated.

[Michael Tully · Accepted solution]

The size you've mentioned is not necessarily correct. If and when there is a crash dump, the dump is actually compressed. The dump itself is usually less than half of the memory and the default location is /var anyway.
In my opinion, if /var is at least 2Gb you
will most likely get all of it anyway. For systems that have *really* large memory you can increase this actual dump space to about 30-40% of memory. This should be enough. This dump area, should also be on the same physical disk as that of your dump.

# lvlnboot -v

[Global Server Operation]

Thanks I have /var at 2 gig. Are you saying that the 4096 of memory is not correct? This is what is on the packing slip that came with the server. I servers are not powered on so I can't verify this.

[Michael Tully]

The memory should be correct. Use these commands to look at the memory on your system.


# echo phys_mem_pages/D | adb64 -k /stand/vmunix /dev/kmem
Multiply the answer by 4096

or if you have the diagnostics installed, this command will lay out your memory completely

# echo "selclass qualifier memory;info;wait;infolog"|cstm

[Global Server Operation]

Thanks!

[S.K. Chan]

It's your choice really, I have had /var/adm/crash setup as a separate FS and the guideline that I used typically .. /var/adm/crash size should be at least half the size of the physical memory (if /var/adm/crash is a separate FS). So far this has always worked for me. I'm taking in terms of "uncompressed". If you decide to keep "crash" dir within /var FS, then make sure you take into account other "potential" log files that are going to fill up /var.

[Jeff Schussele]

Hi,

Michael & S.K are correct as to sizing - although we never go below 50% of memory. I've had 32 Gb systems dump 14Gb+ of files.
We set our crash FS's up on seperate disks & link them to /var/adm/crash. Mainly because we run a lot of large memory systems & have a standard build procedure that takes this into account, i.e. we can't always fit it into /var. Another benefit to not using vg00 for crash is the speed at which the dump can be written. It's inherently much slower writing the crash to vg00 then somewhere else. But we've also found that V-class systems have trouble writing dumps to XP512 arrays unless the V-class is completely up to date on patches & fibre-card drivers. Not sure what the problem was because the same system had no trouble writing them to an XP256. It took HP several weeks to isolate the problem & correct it.

Rgds,
Jeff