March 2005 Archives
March 23, 2005
MATLAB 14 SP2 has been installed on the msprime workstations. Some problems discovered in SP1 seem to be fixed in this newest release. You can try the newest matlab with the command
matlab_r14sp2. The release notes have more information on service pack 2.
On Tuesday, I made MATLAB R14SP1 the default version on the msprime workstations in place of MATLAB R14; some people are having problems with this installation of the newer version. While I find and fix the source of the problems, I have made R14 the default version. Recall that specific versions of MATLAB can be invoked with the commands
March 22, 2005
The msprime workstations have MATLAB releases R13 (ver. 6), R14 (ver. 7) and R14SP1 (version 7.1) installed. The command
matlabnow runs R14SP1 instead of R14. Specific versions can be run with the commands
March 17, 2005
While there is no sliver lining to the dark cloud of yesterday's half-day server outage, there is some good news to report. The first is that no data was lost: all incoming mail was queued for delivery and the data from the dead disk was reconstructed when the RAID array rebuilt itself. The second is that more than 90% of the network boot and standalone linux workstations survived the outage without a reboot - in theory, no program runs would have crashed or outputs written to /scratch been effected by the outage.
Mathserv crashed when a disk failed on Wendesday afternoon and came back on-line at 1:00 am on Thursday once the disk array was rebuillt. Again? you ask. Yes and no. Yes, this is the fourth time this year (and the fourth period since we moved to a new server in the fall of 2003) that we've had Web, email and file-server access go down due to a disk problem; but no, since all four crashes have been due to different causes.
March 15, 2005
The restored back up server caught up over the weekend and is no longer copying files from mathserv (which resulted in mathserv feeling a little logy). Backups of mathserv and all other departmental servers and workstations are being completed by 6:00 am each day so the performance impact is minimal.
March 11, 2005
While the new primary server - formerly the fail-over server - generally performs better the the former primary server did, there have been one- to two-minutes stretches a few times a day when the server was nearly brought to its knees: msprime workstations would start stuttering and web and email responses would be very slow. The source of the problem - a bad table in a research database - was fixed on Tuesday and no more load spikes have been observed.
Our backup server is back online after having disks replaced. Because the primary file system was lost, we are having to prime the backups from mathserv: the I/O activity is keeping mathserv's disks and the HH network rather busy, so the msprime stations will feel a bit sluggish at times (e.g. two to three second delays when opening an application or window), probably until some time Saturday.