cancel
Showing results for 
Search instead for 
Did you mean: 

ulimit -m

Ian Croft
Occasional Contributor

ulimit -m

Hi - I have 2 identical ES40 servers with 4096 Mb of physical memory running V5.1B Patch Kit - Following a recent patch kit BB26 installation the output of ulimit -m shows different results on each server

Server A - 4063808
Server B - 4060408

I have checked the output from sysconfig -q ipc, sysconfig -q vm and sysconfig -q proc and the only paramter different ( Apart from the obvious swap device ) is the per_proc_data_size is different on each server

Server A per_proc_data_size = 1610612736
Server B per_proc_data_size = 3221225472

But I dont think this has anything to do with it.

Does anyone have an ideas why the output of ulimit -m is different on these servers now ?

Thanks in aticipation

Ian
10 REPLIES
Hein van den Heuvel
Honored Contributor

Re: ulimit -m

Bad memory? GH memory? PAL code?

I would use 'vmstat -P' for some details on physical memory


hope this helps a little,

Hein.
Ivan Ferreira
Honored Contributor

Re: ulimit -m

Memory limit cannot be greater than the physical memory. Check how much memory do you have in your systems as described above, with vmstat -P
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Ian Croft
Occasional Contributor

Re: ulimit -m

Thanks for your suggestions so far - I have checked that on both systems and they appear to have the same physical memory - These 2 systems are part of a trucluster and I would of thought given the settings in sysconfigtab being more or less identical we would get the same results. - Any more ideas ?
Venkatesh BL
Honored Contributor

Re: ulimit -m

The 'ulimit -m' gives the sum of all the 'managed' pages displayed in the 'vmstat -P' output.

Ian Croft
Occasional Contributor

Re: ulimit -m

I dont get that though one one server

ulimit -m
4060408

Managed Pages Break Down:

free pages = 355652
active pages = 13580
inactive pages = 30758
wired pages = 50638
ubc pages = 57188
==================
Total = 507816

These dont seem to match or am I missing something ?

Venkatesh BL
Honored Contributor

Re: ulimit -m

The ulimit output is in Kilobytes...

So, 4060408 KB

The 'Managed' pages total = 507816 Pages
So, (507816 * 8192)/1024 = 4062528 KB

Difference: 2120 KB = 2.07 MB

The 'scavenge' space seem to tally with this. So, I think:

ulimit mem size = (Sum of 'managed' pages - scavenge space)

[The formula works for the other server as well!]
Ian Croft
Occasional Contributor

Re: ulimit -m

OK That makes a bit more sense but has anyone any ideas on why the systems are different ? And how I can stop them being different ?
Hein van den Heuvel
Honored Contributor

Re: ulimit -m

I think that as this point you already did enough due dilligence. You saw a *minute* difference, and made sure nothing is dramatically wrong. You know of one param being different (per_proc_data). Declare that to be the cause and move on!

I suspect the the effect on per_proc_data is page table entry pages, multiplies by the number of active processes. If the application is going to be the same, then one of the two settings is 'better':
- allows the right growth
- doesn't waste (modest) memory
Pick that value and move on!

Regards,
Hein

Venkatesh BL
Honored Contributor

Re: ulimit -m

Are you facing any problems due to 2MB variation? I am wondering why you expect the values to match *exactly*.

PS: You may check out http://forums1.itrc.hp.com/service/forums/helptips.do?#28 to learn about the member point system in this forum.
Philip Lawrence_1
Occasional Advisor

Re: ulimit -m

in /var/adm/messages do both systems give the same values for physical memory and available memory at boot time?