Topic: Sluggish network performance- rebooting resolves issue.

[ber]

Linux SME 8 with latest updates. Running in Server/gateway mode. DHCP and Dansguardian running. DC not operating on the local network. There is a windows SQL server running serving databse requests for a membership software.

Hi I'm trying to resolve a intermittent network bottleneck issue for a windows program that is running on 8 workstations. The SQL serves a club membership software and the Network traffic would be described as light-medium.

About once a week the software will run slow when trying to access data from the SQL database. usual display of info time 1-2 secs. When the problem occurs it will take 8-12 secs. The fault is sudden and not gradual.
The only way to resolve it is to reboot the SME server. Rebooting the SQL server doesnt make any difference.

Ive installed sem8admin and bandwidthd to monitor the network traffic. On the network local load- it averages 100-200kbps?? seems minimal/average.
There are no unexpected network load spikes and when the fault does occur; I check the network load logs and its average or minimal. CPU is low and RAM usage is normal. Hardware is good. Nothing in the sme8admin graphs jump out??
Im suspecting that the network pipes are clogged for whatever reason. Where would be the best place to start troubleshooting when the fault does occur.
Thanks to all the volunteers- putting in a donation tonight in appreciation for all the work done- Thank You.

[janet]

ber

Graphs are NOT likely to show you short term issues, you need to use the appropriate network command line monitoring tools.

Other things could be going on eg Browsemaster issues, have you checked settings on both servers & even on the workstations ?
Search here for more info

The only way to resolve it is to reboot the SME server. Rebooting the SQL server doesnt make any difference.

The ONLY way ???
Have you considered stopping certain services to see what happens, rather than rebooting.
You may be able to narrow down the source of the problem better if you stop using the "big gun" reboot approach & get more detailed info, look at log files & settings etc.

[nicolatiana]

Among details required: network card chipset, brand/model of your network hub/switch.

Nicola

[ber]

Hi Two network Cards:

Local: product:   NetXtreme BCM5782 Gigabit Ethernet
vendor:   Broadcom Corporation

External: product:   82557/8/9/0/1 Ethernet Pro 100
vendor:   Intel Corporation

2* 24 port Switches (linked)

- Cisco Systems 3500 series XL, 24 port switch (managed)
- D-Link DES-1024D 24 Port Switch (unmanaged)

[Stefano]

are sql server and w8 client both on lan side?
if so, are them with static ip or dhcp?

[Jáder]

Hi Two network Cards:

Local: product:   NetXtreme BCM5782 Gigabit Ethernet
vendor:   Broadcom Corporation

External: product:   82557/8/9/0/1 Ethernet Pro 100
vendor:   Intel Corporation

2* 24 port Switches (linked)

- Cisco Systems 3500 series XL, 24 port switch (managed)
- D-Link DES-1024D 24 Port Switch (unmanaged)

Everything on HW appears to be VERY GOOD but that D-LINK SWITCH.
If you have a small number of users, could you shut it off for a few days to test ?

Anyways, I'd put all my important computers (8 computers , sql server , SME8 server) on Cisco managed server, and create all possible graphs (zabbix?) of info from switch data (using SNMP). Cisco will tell you what's wrong.
Also take notes about EXACT time when problem happens and search logs for errors (on SME8 and CISCO switch).
And watch CISCO port where DLINK is connected: search for a lot of errors, ...

BTW: Why do you need that D-LINK (I personally hate DLINK, sorry! I prefer to buy LINKSYS, and Cisco agree... they bought LINKSYS)

[mmccarn]

[caveat]
I have not read this whole thread...
[/caveat]

I saw a similar problem on a NON-SME linux server earlier this year.

The cause turned out to be the NIC drivers.

That server had 4 network cards installed.  Two of the cards turned out to work fine, but the other two would work OK for a while after each reboot, then slow down over time.

Something in the drivers for the two troublesome NIC cards seemed to be poorly thought out - those cards were logging an entry to syslog for every packet -- or at least several hundred or thousand logfile entries per second.

[CharlieBrady]

Broadcom NIC hardware and drivers have been problematic:

http://lists.us.dell.com/pipermail/linux-poweredge/2003-January/005936.html

I'd recommend OP replaces the Broadcom NIC with a second Intel one, and reconfigure the server. See if the problem goes away.

Another approach would be to look more closely at the problem. "ethtool -S ethX" will show lots of statistics about NIC status - particularly errors.

[ber]

Fantastic feedback...will report back on finding when i have fully assessed the fau :-)lt and resolved with a permanent fix. Thanks.