Network Trouble...

[Igaduma]

Hi all,

I got an ultra 10 running as fileserver at 100fdx with an external discpack, i'm using samba 2.2.2 and standard NT4 sp6 clients, drivemapping name-to-ip resolution is done via /winnt/system32/drivers/etc/hosts file
All goes well for a few minutes, cpu at 80%, iostat looking ok, no blocking, vmstat seems clean and top seems to report heavy, but acceptable usage & load.
Note that i'm ftping with 5 clients, all at 100fdx.
Then all of sudden *all* network activity stops and the machine can't even be ping-d on the same subnet.
Then the weirdest thing happens:
I type snoop to check the network packets and whatdaya know, the machine comes up again and continues file-serving over ftp & samba goes back on-line.

what the heck is wrong ?

btw, router is configured at 100fdx & so is the machine, not using autoneg.

please help.....i'm silently dying here, went live with oracle db & application, now bottleneck is this problem...each time the disk stops grinding, i type snoop and all continues....

 

[Igaduma]

Anyone ??
<insert begging sound here>

 

[packageman]

What do you see when you do the iostat, vmstat , etc when the system seems to lockup (as opposed to BEFORE it happens)? You might want to check the output of ipcs and netstat -af inet.

 

[Igaduma]

Well, i whish I could, but since last night I copied all static files to each client's local disk and that problem has gone away.

Only serving now &quot;prediction-files&quot; which are dynamic. Seems to be OK now, asked network guys to *confirm* to me everything is at 100FDX. Maybe they reconfigured some ports that where running at half on the router.
Currently it's serving +- 15 people who generate 100k & 500k files each 5-10 seconds (GSMsite simulation application) without any trouble or glitch.
Disabled the wins server that was configured in samba & did a reboot.
Previously I installed it and didn't reboot...couldn't cycle inet, had to continue running like that. Maybe that did it...
Also, the errors seemed to be on socket 4 & socket 12.
I disabled those in /etc/inet.d/services & protocols.
Dunno if that cured it.

L