topic Re: top in Operating System - HP-UX

top

Fadia Almarei — Mon, 07 Feb 2005 02:33:41 GMT

Dear All

i have upgrade my memory for rp5430 server running HP-Ux11i from 2GB to 4GB, i see from the sam and from the dmesg that the memory become 4GB, but when i start the top command it give me that the memory is 2GB what is the problem .

Memory: 2650048K (1291452K) real, 5242888K (3444212K) virtual, 794212K free Pag
e# 1/28

BR,
Fadia

Re: top

Keith Bryson — Mon, 07 Feb 2005 02:57:47 GMT

Hi Fadia

The 'real' memory figure given by top is the same as 'user memory' reported by glance (if you have it) and not actual physical memory of the system. Therefore, this figure doesn't include memory used by buffer cache, the system or any free memory available.

Keith

Re: top

Fadia Almarei — Mon, 07 Feb 2005 03:03:47 GMT

thanks keith
but what is the each of these numbers stand for

Re: top

Artyom Voronchihin — Mon, 07 Feb 2005 03:11:34 GMT

Hello!
Each process reserves as much space in swap as it takes in RAM. So total amount of RAM that processes may use can't exceed the size of swap. To use additional 2GB you have to extend your swap to 4GB or add extra swap partition or set kernel parameter SWAPMEM_ON into 1.

Re: top

Keith Bryson — Mon, 07 Feb 2005 03:25:22 GMT

Hi again

Although Artyom's statement is correct, that won't change the top output. I can only deduce the figures represent the following:

'real' figures represent total user memory (not buffer cache, free RAM or system usage). I can't work-out what the figure in brackets represents though!!

'virtual' figures represent the amount of virtual memory reserved in the system, with the active amount of virtual memory in brackets.

HTH - Keith

Re: top

Fadia Almarei — Mon, 07 Feb 2005 03:44:44 GMT

i have now 4GB memory , and i extend the swap to 8GB by adding another logical volume to the excesting swap which was 4GB,and the total top command is
System: Fin02 Mon Feb 7 10:43:32 2005
Load averages: 0.68, 0.69, 0.74
390 processes: 361 sleeping, 28 running, 1 zombie
Cpu states:
CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
0 1.17 17.8% 0.0% 3.4% 78.8% 0.0% 0.0% 0.0% 0.0%
1 0.20 7.1% 0.0% 4.6% 88.3% 0.0% 0.0% 0.0% 0.0%
--- ---- - - - - - - - -----
avg 0.68 12.5% 0.0% 4.0% 83.6% 0.0% 0.0% 0.0% 0.0%

Memory: 3201720K (1156700K) real, 6294048K (3268712K) virtual, 69812K free Pag
# 1/33

CPU TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
0 ? 4657 apps11i 179 20 76164K 22272K run 0:55 4.22 4.2 5.42 f60webmx
0 ? 3924 apps11i 152 20 540M 114M run 2:17 1.22 1.22 java
0 ? 4832 oracle8i 178 20 723M 9364K run 0:33 1.11 1.10 oracleOA11
0 ? 8639 apps11i 154 20 65548K 12036K sleep 0:17 1.03 1.03 f60webmx
1 ? 8296 apps11i 154 20 12932K 5740K sleep 0:00 0.87 0.87 FNDFSFNDFS
1 ? 4314 apps11i 152 20 255M 42448K run 0:53 0.86 0.86 java
1 ? 7744 apps11i 154 20 80088K 26364K sleep 0:34 0.79 0.79 f60webmx
0 ? 4558 oracle8i 154 20 723M 9268K sleep 1:09 0.64 0.63 oracleOA11
0 ? 1242 root 154 10 3160K 300K sleep 0:38 0.60 0.60 diagmond
1 ? 3923 apps11i 152 20 541M 121M run 0:52 0.59 0.59 0.54 java
0 ? 1221 root 154 20 29520K 6636K sleep 0:12 0.55 0.55 supervisor

Re: top

Keith Bryson — Mon, 07 Feb 2005 04:01:34 GMT

Hi Fadia

Although my explanation is still the same for the top figures, I assume you now have less free memory as some processes are running that weren't running before (maybe they couldn't because you had insufficient swap space).

As Artyom said, swap has to be at least 1xRAM, as the system reserves the same amount of swap that is being used in physical RAM (in the event that the whole of RAM is swapped to disk for any reason).

Keith

Re: top

Fadia Almarei — Mon, 07 Feb 2005 04:07:19 GMT

no i have 8GB swap for a 4GB memory
# swapinfo -mt
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4096 167 3929 4% 0 - 1 /dev/vg00/lvol2
dev 4096 0 4096 0% 0 - 2 /dev/vg00/swap
reserve - 4102 -4102
memory 3147 517 2630 16%
total 11339 4786 6553 42% - 0 -

Re: top

Keith Bryson — Mon, 07 Feb 2005 04:21:47 GMT

Hi

Well that's even better. Something has swallowed-up that extra swap/RAM. It looks as if before there would have been enough swap to reserve equivalent RAM in swap, but that is all. Therefore, if you had used 3.8Gb RAM used, 3.8Gb swap would have been reserved, giving you hardly any swap space left. Now you have added sufficient swap space to allow the system to do swap when physical RAM is low AND reserve total RAM in swap too.

The old rule was SWAP=2xRAM. As RAM has increased in size these days and swapping is a more rare occurrence, the rule tends to be altered to SWAP=1xRAM. Having said all this, I expect that the latter would be for systems with a large amount of RAM. 8Gb isn't a huge amount these days (more's the pity) .....

Do you need any further clarification?

Keith

Re: top

bhavin asokan — Mon, 07 Feb 2005 04:58:44 GMT

hi,

can you post what are the vaules showing against dbc_min_pct and dbc_max_pct when using sysdef command.
i think free memory on your system is very less.
(if you are having glance ,start it,press m ,
see how much it is showing as buffer cache)

regds,

Re: top

Fadia Almarei — Mon, 07 Feb 2005 05:25:22 GMT

this is my output
NAME VALUE BOOT MIN-MAX UNITS FLAGS
dbc_max_pct 50 - - -
dbc_min_pct 5 - - -

Re: top

Keith Bryson — Mon, 07 Feb 2005 05:29:48 GMT

Hi Fadia

This means that the buffer cache on the system may use up to 50% of RAM. Can you check the figure in glance for the current buffer cache usage (option m).

Keith

Re: top

Fadia Almarei — Mon, 07 Feb 2005 05:35:21 GMT

i do not have glance

Re: top

bhavin asokan — Mon, 07 Feb 2005 05:36:32 GMT

hi,

buffer cache of 600MB is enough.change the value of dbc_max_pct to 15% ,it will free a lot of space in your system.
(dbc_max_pct 50% on a 4Gb system means it may take up to 1.5 to 2GB as buffer cache)
note: it is a static parameter ,for changing it you must have to rebuild the kernel.(use sam to change it)

regds,