Server Outage Delays Coursework, E-mail

Several outages and slow-downs in Harvard's central FAS Unix system this weekend plagued students who could not access their accounts, preventing them from sending and receiving e-mail and working on assignments in this, the middle of the reading period crunch.

According to FAS officials, the most serious problems occurred with the "home/01 filesystem," which affects the system access of users whose usernames begin with the letters A through D.

That filesystem was completely unavailable between 7 p.m. and 7:45 p.m. Friday, and again between 7:30 p.m. Saturday and 7 a.m. yesterday. During those hours, affected students could not access anything on their home directory--including the Pine e-mail application, Web pages and class assignments.

"It was serious inconvenience," said Andrew P. Nikonchuk '02, who couldn't access assignments that were in e-mail massages in his account.

Students in computer Science 50, "Introduction to Computer Science I," were the most severely affected. Programming codes which are stored on the server could not be accessed, meaning that affected students were unable to work on a project due early this week. Because of the outage, the due date was pushed back.

"Everyone was trying to do their work at the last minute, and the computers can't handle it", said Daniel I. Jacobs '02, a student in the class. "Everyone had to be given extensions."

On Saturday morning, FAS computer services performed previously planned upgrades on the computer systems. This work was designed to improve the overall speed of the Unix systems. While the upgrade was being made, the systems were slow again, though not interrupted.

The two services outages appear to be related to the upgrades, though in different ways, according to Rick Osterberg '96, FAS's coordinator of residential computing support. Friday's problems occurred while technicians were preparing for the upgrades. The system was shut down because FAS wanted to be sure no bugs would interfere with the work.

Saturday's more serious incident occurred in the wake of the maintenance. To fix the problem, FAS took home/01 offline to avoid data loss.

Osterberg said technicians worked through thenight, backing up all data, flushing out thesystem, and then returning all data to the system.

The systems are designed in such a way so thatwhen one aspect is down, none of the users canreceive e-mail. Thus, all e-mail sent Saturdaynight and early yesterday was not received until 8a.m. yesterday.

According to Osterberg, the fact that bothFriday and Saturday's outages affected home/01 wasprobably coincidence. However, at this time FAScannot definitively say what caused any of thedisruptions. Investigations will continue.

FAS officials said they realized that readingperiod was a poor time to upgrade the system, butthey only received the equipment enabling them todo the work this month.

"We felt that if we did not perform theupgrade, that performance would continue todegrade with increased usage to a point wherethings would be 'too slow' to be acceptable,"Osterberg said.

Minor slow-downs, which are routine afterupgrades, continued yesterday evening. Osterbergsaid future problems with the system are not outof the question.

"There is never a guarantee in this sort ofbusiness, [but] we're fairly confident that we arestable now," he said

Osterberg said technicians worked through thenight, backing up all data, flushing out thesystem, and then returning all data to the system.

The systems are designed in such a way so thatwhen one aspect is down, none of the users canreceive e-mail. Thus, all e-mail sent Saturdaynight and early yesterday was not received until 8a.m. yesterday.

According to Osterberg, the fact that bothFriday and Saturday's outages affected home/01 wasprobably coincidence. However, at this time FAScannot definitively say what caused any of thedisruptions. Investigations will continue.

FAS officials said they realized that readingperiod was a poor time to upgrade the system, butthey only received the equipment enabling them todo the work this month.

"We felt that if we did not perform theupgrade, that performance would continue todegrade with increased usage to a point wherethings would be 'too slow' to be acceptable,"Osterberg said.

Minor slow-downs, which are routine afterupgrades, continued yesterday evening. Osterbergsaid future problems with the system are not outof the question.

"There is never a guarantee in this sort ofbusiness, [but] we're fairly confident that we arestable now," he said