Top command - 4GB ram 1GB ram drive - looks like low mem?

[wilson]

I am not great with Linux, so can someone give me an opinion if this top result looks okay.

The machine is a q6600 with 4gb ram, running ...
slackware 12.1
Asterisk 1.2.30.2
Zaptel 1.2.27
Mysql 5.0.81

Its a single machine running mysql and apache and asterisk and uses eAccelerator.
The is no predictive dialing - all dialing is manual.
All voip based using iax trunk to termination company
users connect via zoiper - ulaw/alaw
Calls over 45 seconds are recorded to a 1GB ram drive
we have about 20 users on at any 1 time

top - 09:20:45 up 1 day, 9:29, 1 user, load average: 0.22, 0.19, 0.14
Tasks: 156 total, 1 running, 154 sleeping, 0 stopped, 1 zombie
Cpu(s): 4.1%us, 0.7%sy, 0.0%ni, 94.4%id, 0.0%wa, 0.6%hi, 0.1%si, 0.0%st
Mem: 2005264k total, 1947588k used, 57676k free, 172032k buffers
Swap: 0k total, 0k used, 0k free, 1002068k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3045 mysql 20 0 512m 103m 3632 S 11 5.3 152:27.99 mysqld
3130 root 20 0 501m 498m 2236 S 1 25.4 27:40.54 AST_VDadapt.pl
4196 root 20 0 35348 12m 4412 S 1 0.6 40:34.99 asterisk
23166 root 20 0 5972 4260 1600 S 1 0.2 0:00.04 AST_send_action
4237 root 20 0 9860 6728 2344 S 1 0.3 10:03.20 AST_update.pl
4246 root 20 0 22208 18m 2280 S 1 1.0 123:44.28 AST_VDauto_dial
14172 daemon 20 0 63812 8640 3564 S 1 0.4 0:00.72 httpd
21329 daemon 20 0 63552 7964 3136 S 1 0.4 0:00.22 httpd
23805 root 20 0 10740 6452 1308 S 1 0.3 0:01.26 FastAGI_log.pl
1 root 20 0 772 304 260 S 0 0.0 0:03.02 init
2 root 15 -5 0 0 0 S 0 0.0 0:00.00 kthreadd
3 root RT -5 0 0 0 S 0 0.0 0:00.10 migration/0

[mflorell]

That looks fine.

[wilson]

Was confused that the total mem looked like 2gb?
and that the free looked too low?

Not really had any problems - other than cross meetme's but I assume this is down to people not logging out.

As we add users at what level should I look at expanding the mem?

Do you think this setup could handle another 10 users? or would I need to get MYSQL and Apache on another machine?

[mflorell]

You're right, it is reporting 2GB RAM.

What linux distro?

[wilson]

Its Slackware 12.1

Someone has suggested that its grabs a chunk of mem and holds on to it in the newer kernels, but I am not sure I buy this.

[gardo]

Looks like the system is 2 gig RAM short. Have you checked if the RAM that you have has no defects? Might be hardware problem. By default, recent Linux distributions (even on 32 bit) can see 4 gig of RAM.

[wilson]

Turns out that there was only 2gb of ram installed
now been increased to 4gb - still only 3gb showing but I think that might be fixed when I recompile the kernel this weeked (64 bit mem addressing)

I now appear to have a problem with a possilbe memory leak?

after 4 hours with under 10 agents the vast majority of the mem is used.

Anyone got any suggestions here?

top - 12:52:55 up 11:17, 1 user, load average: 0.10, 0.17, 0.17
Tasks: 148 total, 1 running, 147 sleeping, 0 stopped, 0 zombie
Cpu(s): 3.3%us, 0.8%sy, 0.0%ni, 95.5%id, 0.0%wa, 0.2%hi, 0.1%si, 0.0%st
Mem: 3299064k total, 2596284k used, 702780k free, 178876k buffers
Swap: 0k total, 0k used, 0k free, 1921152k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3047 mysql 20 0 511m 93m 3420 S 5 2.9 22:18.93 mysqld
3101 root 20 0 27696 10m 4680 S 4 0.3 12:30.49 asterisk
3180 root 20 0 10316 7436 2164 S 4 0.2 3:01.69 AST_VDauto_dial
3183 root 20 0 192m 189m 2088 S 1 5.9 8:04.21 AST_VDadapt.pl
10612 root 20 0 5360 3860 1456 S 1 0.1 0:00.04 AST_send_action
3179 root 20 0 8964 5988 2204 S 1 0.2 0:07.32 AST_manager_lis
3182 root 20 0 7744 4840 2172 S 1 0.1 0:38.80 AST_manager_sen
4349 daemon 20 0 63696 8720 3636 S 1 0.3 0:00.86 httpd
4519 daemon 20 0 63552 7888 3068 S 1 0.2 0:00.80 httpd
22958 daemon 20 0 63808 8632 3560 S 1 0.3 0:02.64 httpd
26455 daemon 20 0 63552 7896 3076 S 1 0.2 0:02.22 httpd
1 root 20 0 772 300 260 S 0 0.0 0:02.00 init
2 root 15 -5 0 0 0 S 0 0.0 0:00.00 kthreadd
3 root RT -5 0 0 0 S 0 0.0 0:00.04 migration/0
4 root 15 -5 0 0 0 S 0 0.0 0:00.04 ksoftirqd/0
5 root RT -5 0 0 0 S 0 0.0 0:00.04 migration/1

[mflorell]

are you doing full recording?

All memory being used doesn't necessarily cause problems. We have seen this at many client sites and Asterisk will reused memory and shouldn't cause problems.

[wilson]

We are doing recording after about 45 seconds.

I think we have call quality problems as a result.

When I restart the server the call quality picks back up - so I am assuming that it is down to this and not connectivity problems.

We use a ram drive, that appears to have reset itself to 1.6gb since the extra mem has gone in (based on df -h). I am going to change this to about 512k this weekend when I recomile the kernel as we mix and move recordings every few minutes.

[mflorell]

Are you using a RAM drive or tmpfs?

what is the peak loadavg on the system?

[wilson]

Total Calls in/out on this server: 0
Total Off-Hook time on this server (min):
Average/Peak channels in use for server: 21.7499 / 53
Average/Peak load for server: 17.4583 / 112
Average USER process cpu percentage: 3.0636 %
Average SYSTEM process cpu percentage: 0.8867 %
Average IDLE process cpu percentage: 96.1238 %

Its tmpfs, do you think this is the problem.
When I rebuild the Kernel I will get this changed if thats the case.

[mflorell]

That looks fine, what zaptel timer are you using?

[wilson]

x100p clone

[mflorell]

How do your zttest numbers look running it for an hour?

[wilson]

--- Results after 5254 passes ---
Best: 100.000000 -- Worst: 99.768066 -- Average: 99.991840

[mflorell]

That 99.7 isn't very good. We stopped using X100p cards almost 2 years ago because they weren't working as well anymore and more and more newer motherboards would just crash when you tried to load zaptel and the x100p module.

What kind of agent phones are you using?

[wilson]

i have a TDM400P, the server is hosted so it will be a job to swap it out, but do you think that will be worth it?

The client phones are zoiper and eyebeam - mainly zoiper

[williamconley]

switch to ztdummy. if it improves, you can either put in a better timing source or not (i use ztdummy most of the time until the server is seriously loaded).

[wilson]

Will switch to ztdummy and see how we get on.

Everything appears to be working well at the moment, which makes me think that the call quality might be more related to our local connection.

The only problem we have now (I have just spotted it) is that some random call recordings are 44 bytes in size both after the mix and both the in and out files. Any ideas on this? do you think it could be related to the timming?

calls are all ulaw and saved as wav