Hi,

I am hoping that some of you may be able to help me.

I have done a search and found a few things, but I am still a bit unsure if there is any solution for this problem.

I know you will probably tell me that this is a feature of Linux and not a bug, but I’m sure that system should not freeze and also use the page file.

The problem that I am having is my Red Hat (v9.0) server is holding on to it’s memory, and this is causing slight freezes / delays in the system. The “Freeze” will last approximately 0.5 to 1 second during which time nothing can be done of the system, anything that is typed is not displayed (i.e. The system does not catch up).

Users are experiencing a number of problems when navigation through the file system, ‘VI’ing’ files and using files via Samba that are stored on this file server.

I have tried a number of things, but don’t seem to be able to stop this “Freezing”

Below is a list of things I have noticed:

vi – If I vi a file it uses 50mb of memory the 1st time I open it. If I then close the file it drops 20mb of that usage and any subsequent file opening will only use an extra 20mb (i.e. every time a file is vi’d it will use 30mb of the system memory that I don’t get back)

Textpad via Samba – If I open a file using textpad (text editor) via a samba share on the machine it will use 20mb of memory, this will not drop off after the file is closed.

Moving data via samba – If I move data to and from the file server via Samba / explorer all of the systems memory is used, this is not dropped off after the move has completed. The only way the memory usage drops off is if the data is deleted from the fileserver.

I have noticed that the system will use about 70mb or 80mb of the page file.

We are running:

Redhat 9.0
Pentium4 2.4GHz
ABIT IT7-MAX2 MB
2 x 1GB PC333 DDR
1 x 40GB Boot disk
2 x 120GB RAID Data disks (RAID on MB)
32MB AGP Graphics card
10/100/1000 NIC

This machine is a file server for 5 Linux computational nodes and also users log on to this server using hummingbird exceed (kterm)

Any help would be appreciated

Thanks in advance

Have you been experiencing this right from the start i.e. soon after the installation ? Or after installing a specific software package/app or upgrading the existing pkgs ?

I don't think we have been experiencing from the start...

Well I say that but the machine had it's boot disk replaced a while ago and it seems like it could have started from then.

i.e.
Old boot disk - no problems
New boot disk - problems started (same model HDD)

I can not guarantee this is when the problems started as the users didn't mention it straight away!!! but it does seem likely.

The only other thing that we have changed on the network is we have introduced a D-Link Gigabit switch, but I personally dont think that is causing the issue.

The users here are telling me that the problem is gettin Progressively worse / more frequent freezes

Thanks in advance

Let's see some hard data. Nothing described is that out of the ordinary - in isolation. What does "free -m" show especially if taken over a period ???.
What about a series of "top" displays ???.

Is the "memory consumption" in the memory, or the cache ???.
How many swap partitions - and how big are each ???.
iowait involved ???.

All sorts of things to look at. Unfortunately there are not any good metrics - most of the analysis must be by implication.

Hi,

Thanks for your reply...

Please see information below

Results from 3 free -m commands

........total....used....free....shared....buffers....cached
Mem:.2016....1996.....19.......0........179.......1707
-/+ buffers/cache:110.....1906
Swap:2000....72......1927

---------------------------------------------------

........total....used....free....shared....buffers....cached
Mem:.2016....2002.....14.......0........179.......1713
-/+ buffers/cache:109.....1906
Swap:2000....72......1927

---------------------------------------------------

........total....used....free....shared....buffers....cached
Mem:.2016....1995.....21.......0........179.......1696
-/+ buffers/cache:118.....1897
Swap:2000....72......1927

Results from 3 top commands

11:32:15 up 5 days, 20:11, 4 users, load average: 0.34, 0.21, 0.14
89 processes: 87 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 1.4% user 0.6% system 0.0% nice 0.0% iowait 98.0% idle
Mem: 2065048k av, 2042500k used, 22548k free, 0k shrd, 183108k buff
1577368k actv, 0k in_d, 35564k in_c
Swap: 2048276k av, 74644k used, 1973632k free 1730008k cached

-----------------------------------------------------

11:33:54 up 5 days, 20:12, 4 users, load average: 0.13, 0.17, 0.13
90 processes: 88 sleeping, 2 running, 0 zombie, 0 stopped
CPU states: 1.4% user 0.0% system 0.0% nice 0.0% iowait 98.6% idle
Mem: 2065048k av, 2042492k used, 22556k free, 0k shrd, 182988k buff
1577496k actv, 0k in_d, 35548k in_c
Swap: 2048276k av, 74652k used, 1973624k free 1730136k cached

------------------------------------------------------

11:35:16 up 5 days, 20:14, 4 users, load average: 0.27, 0.22, 0.15
90 processes: 89 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: 1.6% user 1.2% system 0.0% nice 0.0% iowait 97.1% idle
Mem: 2065048k av, 2043776k used, 21272k free, 0k shrd, 181988k buff
1575436k actv, 0k in_d, 36828k in_c
Swap: 2048276k av, 74724k used, 1973552k free 1730892k cached

---------------------------------------------------

We only have 1 swap partition that is 2GB in size (was set up to be double the size of the RAM, but we have recently doubled the amount of RAM in the machine to 2GB

I am sorry, but I am not sure what you mean when you ask if iowait involved ???

Thanks in advance

On the face of it you don't seem to have a problem.
I presume you are on a 2.4 kernel ???. 2.6 introduced a "swappiness" sysctl to give you some flexibility in balancing the usage of storage to better suite your (i.e. the administrators) desires.

I don't know that you have a lot of options if you are on 2.4.

[edit] forgot about iowait - have a look at the top display(s) - you don't have an issue (0.0%)

Yeah I am running 2.4.20-8

Do you think that upgrading to 2.6. would be worth trying?

I couldn't see anything wrong with the outputs from top etc... but the system is still freezing!!

I'm the wrong guy to ask - all my systems are at 2.6, and except for "packaged" distros like Ubuntu or Mepis, are at 2.6.15 or later.
No (stable) servers here buddy ...

There are certainly benefits - and I don't agree with those that say the alleged stability of 2.4.x has been lost. For your situation you might want to cruise by and have a look at something like Centos (RHEL for the masses).
I happen to think swappiness might help relieve your situation - presuming it's not just something like bad DMA settings. Doesn't look like it - bugger all I/O in that (short) snapshot.

Hi,

Thanks for the reply's. I have read the info that you have suggested.

Could someone please take a look at the following /proc/meminfo file and tell me if it looks normal

...........total:........used:........free:.....shared:..buffers:....cached:
Mem:..2114609152 2096459776 18149376 0 223342592 1740263424
Swap: 2097434624 75329536 2022105088
MemTotal:.......2065048 kB
MemFree:...........17724 kB
MemShared:..............0 kB
Buffers:............218108 kB
Cached:..........1691288 kB
SwapCached:........8188 kB
Active:............1544660 kB
ActiveAnon:.........14424 kB
ActiveCache:....1530236 kB
Inact_dirty:.........10044 kB
Inact_laundry:....329400 kB
Inact_clean:........39840 kB
Inact_target:.....384788 kB
HighTotal:........1179584 kB
HighFree:..............6672 kB
LowTotal:..........885464 kB
LowFree:............11052 kB
SwapTotal:......2048276 kB
SwapFree:.......1974712 kB

Thanks

Barry..

The above output likewise the earlier one you posted shows that the system is really not actively using that much memory, as as much as ~ 1700mb is cached.

There should be no reason whatsoever to worry about swap with 2gb of ram.

Like suggested, upgrading to 2.6 kernel might be a good choice. Actually I would do it anyhow already because security etc. reasons. And never ever have experienced stability problems with 2.6. kernels.

You might want to track the issue with e.g. 'vmstat -an 2' which gives a better outlook to for instance io loads. However, my guess is that the problem is not really directly related to memory but to some software issues (e.g. what about the Samba itself?).

The load on both the CPU and RAM are pretty light - as syg00 mentioned "free -m" is a great way to determine how your RAM is being used. The "+/- buffer/cache" is what matters, and shows you what is actually really being used.

If these issues started to occur after a new disk was installed, maybe it isn't using DMA. What are the results of
(assuming your boot drive is /dev/hda)

You'd want to see "using_dma = 1(on)", and more info can be found here

Also, I'd suggest checking how full your drives are, sometimes a system will hiccup if the amount of free space gets too low. Good luck with it