workstation lockup

[xiangli]

I have a B-180 (Ver 10.20 )work station run two X-session login in remote sever via cisco4700 router, it will lockup after someday, doesn't response any keyboard input, it take morethan 10 minute to wakeup screen after it is in power save mode by move the mouse, the mouse pointer can not be moved,can not be access via LAN, it only send serveral packet within few hour, the only way is power off it.
I bring another one (same configuration and appliaction, never lockup) from other site,
sofar, it have not such problem, but I noticed it became slower or slower, it take very longer and longer to reply ICMP ping, if you disconnect lan cable and then put is back, the speed will resotred, I find usally there are one packet take much longer time(500 times) within 5 or more packet, the remain packet act normal, I belived if don't reconnect cable ,it will lockup at last.
thank for any suggestion

[Michael Tully]

Hi,

Sounds like you NIC card is experiencing a problem with auto port negotiation. Do you have Half or full duplex? What is the setting on the switch?

If you run 'lanadmin' and look for the statistics on late collisions, this will most likely tell the the problem on the above, if there are quite a number.

Michael

[xiangli]

it is half half-duplex card, speed is 10M, its's 10/100M card as labeled, but my HUB support 10M only.
I have check the statistic vai lanadmin,the late collisions value is very low, but the defreed transmission number is incresing.
but currently the performance is quite well, I still don't know which action will lead the degread of speed, only thing is sure, if it became slow, it will slower and slower untill you reconncet the LAN cable.
thanks

[Michael Tully]

What setting do have in your switch? It will make a difference if it is set to auto/full duplex and the card using half duplex, even if the late collisions are low. Many slow ftp problems have been solved with this same problem. One other thing I would like to ask is, what patch level do you currently have?

[xiangli]

the latest status is the WS is lockup as the one replaced,it means it maybe nothing about the machine, because this machine work quite well in other place, another possible reason is router, but there are some other IP device(my laptop ex.)all other staff act normal meanwhile.
for info. you need
there are only one dumb HUB, no switch avaliable,for patch, they just apply basic patch.

there are serverl WS in this ring network,all these is same(both software and hardware),they login two remote server.two of them is ok, two will lockup accedentily.


the below text is I find in HP site, it sound helpful for me, but I still don't know how to change the setting,router or WS.

A5346115/TRAK-ACTIVE/English] A5346115 Cisco Router: Message to increase LLC timers; recommendations Problem Description I have a Cisco 4500 router with dlsw. I am receiving a message to increase LLC timers. Performance is slow. Should I increase the llc2 t1-time? Configuration Info Operating System - HP-UX Version - Hardware System - HP9000 Series - Solution Yes, the error indicates is that the T1 timer (ack timer) popped. Increasing the ack timers (T1) should decrease the likelihood of a failure. LLC Timer Recommendations Using SRB Over Slow Serial Links ---------------------------------------------------------- The following recommendations are for timer settings on the hosts in a remote source route bridged environment: T1 - 1 to 2 seconds is the normal default. This should be greater than the total round trip delay time that the frame might encounter within the sending node, the network, and the receiving node. On slower WAN links (< 64K or < 256K on heavily used links), T1 should probably be 3 to 3.5 seconds to reduce retransmissions and connection timeouts. Ti - This value is recommended as being at least 10 times the T1 timer. Try values around 30 for best results. You usually don't need to change this setting. Ti expiration causes an RR to be transmitted, and then a T1 timer is started against the RR. T2 - This normally should be between 80 and 256 msec, and must be less than T1. This is not generally tuned. If it is too long, it could hurt response times and cause unnecessary retransmissions. Notes ----- In a locally acknowledged environment, it is not usually necessary to tune T1, Ti or T2 timers unless a frame is traversing numerous local bridges or routers. T1 = how long a device waits for a reply to a transmitted frame. Ti = inactivity timeout. This is like a session keep alive timer. If a link station hasn't heard from a device with which it has a session for the length of time defined by this timer, the link station will send an RR to see if the device is still reachable. T2 = acknowledgment delay timer. This value defines the maximum time after receiving a packet before responding. The above information is from http://www.cisco.com/. xx