Recently in System Notes Category

Two people reported that their workstations froze with a blank, white screen. In each case, the computer wasn't frozen - only part of the display layer was; all applications were still running in the background. If this happens to you,you can unlock the screen like so ...

1. press alt-ctrl-f1


2. login


3. enter killall -9 kwin; exit


4. press alt-ctrl-f8

The problem is related to a bug in the KDE desktop environment, but I don't know what triggers it. If it happens to you more than once, I suggest you try using the Gnome desktop environment until I sort things out; you can choose an alternate desktop from the menu on the login screen.

The web browsers are particularly sensitive to the workstations losing their server connections now and then - while other applications coast over the bumps, the browsers often freeze.

I highly recommend using opera until Monday - while it still freezes sometimes, it has a much better recovery process and will generally just let you pick up just where you left off.

Things should be back to normal on Monday.

Some people have found that they cannot ssh to redpine (used for SAS). The solution is to login to mathserv and reset your password with the command passwd (you may re-enter your current password).

It appears fairly certain that the second disk really did fail before the replacement for the first failed disk could be built into the array. The file server, workstations and email will be down all day while I replace the disks and recover from files from backup.

Once recovery is under way, I will make a final attempt to recover data from the old disks which may mean that no data is lost. If that fails, any file changes or mail received between 4am and 10:15 am will be lost.

We are now able to backup using the normal windows: four weeks' changes kept for mathserv; two weeks' changes for all other systems.

Bluespruce, which crashed last week, is not coming up anytime soon: the problem is serious but so fair difficult to pinpoint. The system will likely be down for at least a week or two. I am looking into making another compute server available during the interim.

I recommend that you turn computers and monitors off over the holiday given the risk of power outages and the power spikes that often accompany them. It would not be completely paranoid to unplug your equipment if it plugs directly to a wall jack.

Of course, computers which are turned off cannot be accessed remotely and will not be backed up. Note that most graduate students and post-docs have home directories on the server, which should not lose power.

Mathserv stopped allowing new connections for about fifteen minutes shortly afternoon, which led to odd behaviors on the workstations and loss of mail connectivity. Everything is OK again as of 12:20 pm.

The Math and Stats web server stopped serving PHP pages at about 10:30 this morning; it was OK again at 11:05.

A few notes about your linux workstations ...

You are mostly likely sick to death of hearing about the new early start to DST, scheduled for this Sunday. I won't go over it here in any detail. In brief:

The servers and RHPCS-managed workstations* are taken care of.

You should check the time on your Windows or OS X system on Monday - if the time is off, you can find information and instructions here:

• About Daylight Saving Time changes in 2007 from Apple
  
• Daylight Saving Time Help and Support Center from Microsoft

The Math & Stats servers, linux workstations and many office Macintosh systems are backed up nightly. You can check the backup status of your workstation on this page.

Note that most of the linux workstations and servers use the central server for storing home directories and much of the application software.

Two graduate students experienced problems with KDE recently: the application menus disapppeared and they could no longer open files by double-clicking on then in konqueror or on the Desktop. The source of the problem is still a mystery, but we do know how to fix it.

Mathserv was extrememly slow between 7:00 pm and 11:00 pm on Wednesday evening, often to the point of being unresponsive; this affected email, web and msprime workstation services. I rebooted mathserv at 11:00 in order to let it shake itself out after spending an hour isolating the source of the problem (during which hour mathserv was sometimes just fine and sometimes miserably slow).

For those who are interested: mathserv was swamped by a huge file transfer. It should not have reacted quite so badly as it did, so I'll be investigating further.

Mathserv and the msprime workstations were very slow for about ten minutes at around 12:30 pm today due to a massive network load. Email and web access will have been intermittend during this period. Everything should be back to normal as of 12:35.

Mathserv network services failed Saturday at ca. 3:30 am; I rebooted the system at 7:45 pm and all services were back up shortly thereafter. I will look into the reasons for the failure on Monday.

Note that mathserv did not actually crash so there was no data loss. Any mail sent to math accounts will have been queued and should have been delivered by now.

We are experiencing intermitent problems with ssh'ing to some campus servers, including mathserv. Mail and web are not affected.

Mathserv was rebooted half an hour later than announced; the reboot corrected a memory problem on the server. The reboot was delayed in order to allow the nightly backup to finish.

Four people report difficulties connect to off campus web sites from computers in Hamilton Hall on Thursday. I have contacted UTS; please let me know if you have experienced similar problems.

4:20 pm: the problems have been resolved.

Mathserv was down for ca. fifteen minutes this afternoon; it's back up as of 3:10 pm with no apparent loss of data.

You will sometimes see references not just to mathserv but to mathserv1, mathserv2, mathserv3 and mathserv4. What does this mean?