Our unit is a medical institution. The entire hospital network consists of an HP server with win2000 server version and dozens of terminal PCs with win98 as the platform. The database system uses Microsoft's sql 7.0 version, and the development tool is Sybase System's powerbuild 6.0. In the past year's operation, there have been two weird "faults", which almost caused the system to be paralyzed. Here, the detailed process is described as follows, with a view to enlighten everyone.
"Fault" One: After the system was running for a few months, the operation speed of the entire management system became very slow. Whether it was registered or price-priced, the reaction was slow, which made the whole hospital unable to perform normal work. Observing the server, I found that the hard disk lights often flashed, and the system performance monitoring in the task manager was opened. The cpu occupancy rate was high. Clicking on any program icon, the response of the program is very slow, and the refresh rate of the screen seems to be Performing cartoons. The first thing we thought of was a virus infection, so I spent nearly 10,000 yuan to buy a network version of a famous anti-virus software, and I found nothing after scanning the entire network. So suspected that it was infected with some unknown virus, and the comrades of the software company could not solve the problem again. Therefore, it is suspected that the server hardware is faulty. Please ask the HP company personnel to check, cpu, memory, etc. are no problem. In the near future, suddenly I think there may be problems in the configuration of win2000. So, click "Start"→"Program"→"Administrative Tools", and then check each management item under the column, and finally find the problem. Originally, under the "Event Viewer", "Application Log", "Security Log", "System Log" logs, etc., were all filled with records. Some of them have nearly a thousand records. We suspect that when win2000 is running, in order to "ensure" security, it is possible to read information in multiple log files at any time, and compare it with the current situation. What to do next, this will greatly increase the system's resource consumption. Moreover, once there are multiple erroneous records in the log, the situation is even more serious. So, after we backed up, we emptied all the log records, restarted the server, and the entire network was finally rejuvenated.
"Fault" II: After the system resumes normal, accept the last lesson, we periodically clear the log file, and open the system performance monitor to check the occupancy of the cpu at any time. But the strange situation has appeared again. Sometimes, when the entire network is not doing any work, the hard disk light flashes again and again, and the cpu occupancy rate reaches 100%. This situation automatically disappears after about one minute. Unfortunately, if the outpatient or inpatient department of the hospital has a task to do at this time, the response speed of the whole system will be slow as a snail, and it will not work properly. We will clear the log every day according to the last solution, but it will not be solved. problem. After careful observation, we found that this phenomenon occurs regularly every 16 hours, so it is suspected that the plan task similar to win98 is in the ghost, but in win2000 can not find the "plan task" this management item, nothing to do, Going to the log file again, inadvertently found in the system log, there is a pair of "error" and "warning" records that repeat every 16 hours, so the light is on, open "error" At the first glance, the content is " This machine is the pdc in the forest root domain. Use the net command net time /setsntp: to configure synchronization from an external time source. "The error code is 62; "Warning" The content of the record is: " Due to repeated network problems, the time service has not found a domain controller synchronized with it. To reduce network traffic, the time service will take 960 minutes to try again.
There will be no synchronization during this period and the network connection will be restored in time. Accumulated time errors can cause some network operations to fail. To tell the network connection that it has been restored, it should be resynchronized. Execute "w32tm /s"" from the command line. The error code is 64. At first glance, isn’t 960 minutes exactly 16 hours? Since the network of my organization is an internal extranet network, and the software system takes the server time, the network time synchronization service is not needed. The problem of stopping this service will be solved. After searching, we finally found its home. Click "Start"→"Programs" →"Administrative Tools" →"Services". Find the windows time project, double-click it to modify it, stop the service, and change its startup mode to manual. Since then, the 16-hour intermittent "failure" has never appeared.