Completed in October 2007

From Nuclear Physics Group Documentation Pages
Revision as of 15:51, 12 November 2007 by Steve (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search
  • The computer that was once improv, then quark, is now feynman. Jim Williams had the name quark first.
  • Removed the ancient kernels from symanzik's boot partition, because they were taking up space needed by kernels from up2date
  • Figure out how to password-protect a webpage for Silas. He hosts from his personal space on nuclear.unh.edu http://httpd.apache.org/docs/2.0/howto/auth.html#gettingitworking
  • pepper, taro got kernel updates, but we have to wait on some users to finish some long jobs. lzana says that he'll be finished around Tuesday (9/25/2007), so we should send out an e-mail on Monday to inform others. Tenetatively schedule the reboots for that Friday. He's still running his jobs. New estimate for completion is over the weekend. Done on 9/28
  • Jalapeno was hung when I came in, so I took the opportunity to boot it with the latest uniprocessor kernel. Let's see how long it can last with this. If it hangs again soon, then the issue probably isn't the SMP kernel. "Found" a newer SMP kernel, but it panics on boot. Tried the SMP again, but it made it to startup. Let's see how long it goes this time. It could be a power issue (e.g. Taro). Hung up again this morning (8/22). Let's look into the power angle. Hung up last night with the single-processor kernel (8/24). Hung sometime between the mornings of 8/26 and 8/27. Restarted this morning (8/27) with the default kernel. Was hung on 9/03 when the backup script came by, judging by the email records. Stopped by wed night, noticed jalapeno panicked. Copied down everything visible on-screen. Maybe we can use this to narrow down what's happening here.... (9/12, 7:40) Unsolved, just tidying; todo list has another jalapeno entry
  • Removed the "delete" set of fake users from the database
  • Updated clamav on roentgen so that the logwatch e-mails aren't filled with warnings about it being out-of-date. Had to create a symbolic link ln -s libclamav.so.2.0.8 libclamav.so.1 and slightly modify the /etc/clamd.conf file to get service clamd to start Turns out that clamd wanted libclamav.so.1 because it was an old version. Updating the clamd package fixed everything, and let me revert clamd.conf to its proper form. I wonder what led to them deciding to have two seperate packages if one depends on the other…
  • The latest kernel on pepper doesn't have a SMP version? Had to go back to the second-most recent. Using most-recent version now; it just wasn't automatically added to the lilo menu.
  • Fixed some things warned about by UNH's "Nessus":
    • snmpd (mainly used for monitoring by cacti) on pepper was accepting the default community for v1 and v2c. /etc/snmp/snmpd.conf was changed to be more like einstein's and to not accept the public community. snmpwalk -c public -v 1 localhost and snmpwalk -c public -v 2c localhost return with "No response" so the changes seem to have worked.
    • Issue with named on tomato is mysterious. Removed some Aaron chaff, but it's otherwise identical to einstein's /etc/named.conf
  • Wilson can't download http://physics.unh.edu/wheel/custom/rhel-4/i386/repodata/repomd.xml and get updates from the npg-custom repository. wget says that a connection is made, but gets a 404 File Not Found error. I'm not sure that it's a settings issue, because blackbody is able to successfully download the file with wget and Opera. Bigger trouble: Parikshit rebooted wilson while I was trying to solve some interdependecy problems, and so "e2fsprogs" was uninstalled. This package contains programs dealing with ext2 filesystems, including fsck. So, when wilson tries to boot, it wants to do a check, but can't find fsck, and just gives up booting. I tried using the backtrack CD, mounting the LVM, chrooting it, and using yum, but it's not finding e2fsprogs in the two repositories that I added. This could probably be fixed if the http issue is figured out. Fortunately for Parry, wilson dual-boots Windows, so he can run that in the meantime. Where is wilson anyways? It's the computer nearest the window in 303. Set Parry up with ennui for now. Wilson is currently in 202. My plan is to just install CentOS 5 on it.
  • Set up Jim William's dual-boot. I might need to come in after midnight to catch him, considering his scheduling... Installed CentOS5, takes forever on system message bus on boot. Leaving it overnight to see if it fixes itself. Nope. Odds are it's LDAP. It had some major errors in the ldap.conf files (BASEDN is dc=physics,dc=unh,dc=edu not dc=einstein,dc=edu,dc=edu). Still hanging on message bus, though It seems to be unable to connect to NPG stuff, but only in certain ways. It can ping einstein and ssh into it, but can't scp or be ssh'd into. It can reach any website that I've tried except for physics.unh.edu. It was having ennui's MTU problem. Maybe this issue is caused by that port on the switch?
  • Got sick of this so I created a new account for myself (uid "steve"). Steve can't log into roentgen. Don't appear in getent passwd | grep mccoyst, but that's the case on several other machines that I can log into, such as einstein. However, getent passwd mccoyst does return my info on einstein, but not roentgen. ldapsearch -x '(uid=mccoyst)' returns me on roentgen. Fake user "fdelete" had my uid as its gid, so I finally deleted it and the other deletes "just in case". They still show up in roentgen's getent passwd, but not ldap searches, and I was able to log in directly as jdelete and remotely as ndelete… Bwah??:
 root@roentgen:root>luserdel fdelete
 User fdelete does not exist.
 root@roentgen:root>userdel fdelete
 userdel: error deleting password entry
 userdel: error deleting shadow password entry
 userdel: error removing group entry
 userdel: error removing shadow group entry
 root@roentgen:root>id fdelete
 uid=4239(fdelete) gid=4235(fdelete) groups=4235(fdelete),4199(farm)
  • The command-line version of up2date seems to be broken. However, the GUI version works well. Fixed with some RHN licensing