Solaris

Hi All,
using top command I have the below status

Code:

load averages:  3.68,  3.20,  2.67;                    up 0+07:01:24                                                13:13:16
282 processes: 275 sleeping, 3 stopped, 4 on cpu
CPU states: 87.9% idle,  6.9% user,  5.2% kernel,  0.0% iowait,  0.0% swap
Memory: 64G phys mem, 1791M free mem, 63G total swap, 62G free swap

I am noticing the free memory is increasing to 3000M approx when it reaches 1500M.

We had 27GB free memory when restarted the server,But it continuously decreased to reach 1500M and keeping the free mem between 1500 to 3000.

Thanks,

I notice you're not using most of your swap; so I don't think most of that 63 gigs has been leaked or anything. I don't see how much cache you have from that, but suspect a pretty good chunk is cache.

It's kind of random, and determined by disk access. When a process explicitly frees memory or quits, the memory it used goes back into the free pool. If you read from a spot on disk that hasn't been cached, enough free memory is converted to cache memory to cache it. So it's the rate your processes are allocating and freeing memory, versus the rate you're asking for things on disk that haven't been cached lately.

Since cache memory is as good as free memory -- the OS will give it up at need -- there's nothing wrong with that. Just the OS hedging its bets in case it will be handy later. Memory that sits around doing nothing is useless anyway.

Some systems allow you to configure how much memory should be allowed to become cache and the like, but the same system still underlies it.

@]megh: What Solaris release (cat /etc/release) and what file system (UFS and/or ZFS) are you using ?

Both can have a huge impact on what top, which isn't recommended anyway on Solaris, is reporting.

I would suggest using Solaris specific commands to monitor memory usage.

Code:

vmstat (free and sr columns)
prstat -Z
swap -s
echo ::memstat | mdb -k

Code:

ipcs -m

may also shed light on what is using memory.