Thanks for the advice Peter! The customer has made some furher investigation with locked-up Max20's and seems that the box is alive, but the Ethernet port is jammed. Removing and re-inserting the cable does not help. Nor does clear int or shut, no shut in an adjacent Cisco. Are there any commands to clear the Ethernet interface in Max20? The problems with 8.0.3 were in OSPF. Now the customer has upgraded some boxes to 8.0.3 with the same config as 7.10.6 (RIP). -Mikko At 09:35 03.11.2000 -0800, Peter Lalor wrote: >>From: Mikko Lahti <makko at cygate.fi> >> >>Our customer has a problem with Max20's crashing frequently. Does someone >>have experience of similar problems before? Or, can someone give advice >>how to find out what makes the boxes to crash? > >IME, the following makes the DSL Max 20/DSL Terminator 100 unstable: >1. Flapping profiles (cause crashes). >2. RADIUS Authentication (various problems, including crashes). >3. OSPF (can cause Ethernet to go deaf). >4. Physical disconnect of DS-3 (fixed in 8.0.3). >5. Turning them on. ;-) > >>They have some dozens of Max20's, equipped with either one or two SDSL-16 >>cards. Most of the Max20's crash frequently. After a crash, a power >>reset is >>needed to get the box up and running again. Some of the boxes crash very >>frequently, up to 4 or 5 times a day. The software level is 7.10.6. > >They almost always come up again after a crash, IME. Only once or twice >I've seen routing be flaky and I had to reset again. > >Have you tried fsave, nvramclear? > >>Things that seem to affect the probability to crash: >>- - Number of SDSL cards in the box. >> Boxes with two cards crash more frequently. >> However, boxes with only one SDSL card crash also, >> but not so frequently. >>- - Number of lines used on those cards. >> More lines in use => more probable to crash. > >We only use one with a single SDSL card. It is totally stable, running >8.0.3. The rest use T1 and/or ATM DS-3. > >>The Max20's are configured to: >>- - PPP >>- - RADIUS authentication >>- - customer subnets are parts of C-class networks > >Try migrating a single box to local profiles (you can still use RADIUS for >accounting). Does it get better? > >>In the beginning, only static routes were used. Later, RIP2 has been >>used. This change seemed to have no effect on the crashes. > >RIP is not a problem. > >>Syslog does not give any information about the crashes. The boxes die >>without giving any syslog messages. Also, there's no information coming >>to a RADIUS server at the time of a crash. Is there a way to get more >>information, what is happening in a Max20 at the moment of a crash? > >What does the fatal error log show? Diagnostics>'fatal' > >>The customer has tried 8.0.3 software, but it had even more problems. > >Such as? It's better than all previous releases for me. >-- > >Peter Lalor Infoasis >plalor at infoasis.com http://www.infoasis.com/ > >"Where's my burrito?" -- Homer ++ Ascend Users Mailing List ++ To unsubscribe: send unsubscribe to ascend-users-request at bungi.com To get FAQ'd: <http://www.nealis.net/ascend/faq>