size of /var/adm/crash

[Popy]

I have 24GB memory, 16 GB swap/dump and got only 8GB of /var/adm/crash.

Do we really need a /var/adm/crash exactly same size of my dump space ? or 8GB /var/adm/crash is sufficient ?

[Ramesh S]

Hi

It depends upon crash configured on the system

# crashconf -v

Have a look on Total pages included and dump space.

Best Regards,
Ramesh

[Don Morris_1 · Accepted solution]

That rather depends on what your kernel is doing if you happen to get a crash and what version of HP-UX you're referring to [newer releases will compress the dump so you need less space, etc.].

crashconf(1M) will give you a basic overview of your dump needs at any given time (including if your system supports compression, etc.). Certainly if it states you need more than 8Gb now you're in trouble... otherwise, we're in Magic 8 Ball territory. There are certainly cases where I could easily imagine an uncompressed dump being in the 22Gb range [problem in I/O space where buffers are allocated and never freed, for example... in-use kernel dynamic memory is dumped for obvious reasons -- if more than 8Gb is in use (and the dump isn't compressible) you're going to have an incomplete and likely useless dump.

If you don't think you'll ever actually need the dump space because you either won't crash, you know your consumption will almost always be in User Space [if you're running a big Oracle SGA that consumes most of the memory and memory locks it, that's a reasonable assumption, for example] or you just don't care to have the dump be analyzed if you ever actually get a crash, you can keep the dump space small. Otherwise, I'd at least make it the 16Gb.