sasjfp@unx.sas.com (Jeffrey L. Phillips) (11/30/90)
Greetings: We have been having some trouble since we recently converted about 120 sr9.7 machines to 10.2 and I am looking for help. Most of these problems HPollo cannot reproduce, so, therefore, I must not actually be having these problems (I'm a chronic lier :-)). 1. /dev/sio1 is ALWAYS locked on our 10.2 machines. HPollo has several print server machines that they cannot reproduce these errors on. We have to remake the sio devices everytime we start the print server or access the /dev/sio1 line. Kermit often fails also. The only fix we have so far is to prepend an '/etc/mkdev /dev sio' in front of any commands that use sio lines. This is not convenient, for only root can exec. /etc/mkdev. 2. Alarm server sends an error alarm everytime most users log in. The alarm (forgive me, I don't have the actual text) is the same alarm you get if you are logged in to two machines (in other words, two alarm servers running). It's basically the same problem as above. The user's msg box, user_data/alarm_server.msg_mbx is locked as if it were in use. This causes problems when trying to do a send_alarm -u (a send_alarm -n still works since that accesses /tmp/alarm_server.msg_mbx). I have yet to figure out why this file remains locked when the one in /tmp is not. Our only fix is to kill alarm_server, remove the file (with a -du switch), and restart the alarm server. After that it works fine for about a week, then acts up again. 3. About 50 machines or so are getting the sfcb hash table mutex lock problem on a random basis. I am sure that something is causing it (most of these users use GNU emacs quite a bit. That is listed as a problem in the patch tape doc). We installed the patch, but are still getting problems. I am not running anything strange, nor is my machine very busy, but I still have to reboot my machine every morning before I can log in because of this. We have no fix for this at all. I realize that I have unloaded here, but any suggestions would be well appreciated. I don't have too many strands of hair left to pull out. Thanks in advance. -- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ "Just because I don't know what it means doesn't mean I'm lying." Jeff Phillips Email: sasjfp@dev.sas.com
obrennan@CC3.CC.UMR.EDU (obrennan) (11/30/90)
Greetings: We have been having some trouble since we recently converted about 120 sr9.7 machines to 10.2 and I am looking for help. Most of these problems HPollo cannot reproduce, so, therefore, I must not actually be having these problems (I'm a chronic lier :-)). 1. /dev/sio1 is ALWAYS locked on our 10.2 machines. HPollo has several print server machines that they cannot reproduce these errors on. We have to remake the sio devices everytime we start the print server or access the /dev/sio1 line. Kermit often fails also. The only fix we have so far is to prepend an '/etc/mkdev /dev sio' in front of any commands that use sio lines. This is not convenient, for only root can exec. /etc/mkdev. -- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ "Just because I don't know what it means doesn't mean I'm lying." Jeff Phillips Email: sasjfp@dev.sas.com -------------------------------------------------------------------------------------- We had the same sort of trouble. I believe the problem was that the /sys/node_data/dev directory wasn't built correctly. It happened to be the node with the software source area (hard-linked). I believe we ended up using mkdev for the entire devices directory which solved our problem. It's been a while, so I hope I'm remembering correctly. Hope this helps... Gerry O'Brennan Programmer/Analyst II Computing Services University of Missouri - Rolla ------------------------------ obrennan@apollo.cc.umr.edu c0022@umrvmb.umr.edu ------------------------------