How do you fix (repair) non-working programs for users?

Question

System administration

Good afternoon. The approach of local admins is interesting, who are either now part-time engaged in user support, or were engaged when they were enikeys. How do you deal with hanging processes, for example? . Recently, my user constantly hung the process when opening any pdf file in adobe acrobat, when submitting it for printing. Before that, I solved similar problems by stupidly reinstalling the software. This time, the reinstallation did not help, I did not find anything intelligible in the logs. As a result, I drove the dump through windbg for 2 days, and as a result, with grief, I found out in half that I was hanging the splwow64 process, one driver that came from nowhere. Installed the latest version of the driver and it worked. 2 days for 1 request from a user in my office is just too much time. The bottom line is a moral dilemma. Repair software by reinstalling, on the one hand everything works fine, on the other hand, unprofessional. I want to know who does what, if no obvious reasons for the breakdown were found in the logs?

Answer 1

How do you fix (repair) non-working programs for users?

Blue electrical tape, shamanic tambourine, and most importantly - disk defragmentation!

I want to know who does what, if no obvious reasons for the breakdown were found in the logs?

Rolling back the system to a workable configuration saves in such cases, and there is no need to look for a breakdown.
A balanced approach to choosing software and setting rights also helps a lot.

Answer 2

Popular software - by Google (in extreme cases, a detailed description of the problem in the developer's bugtracker), self-written from the description of the incident to the developers with a description of the reproduction of the problem and all the details, or a debugger with a bible. Paid software - chiseling those. support and poking in sla.
It is important to understand the difference between an incident and a problem. If something crashed once and the reinstall helped, the incident is over. If the employer agrees to pay for picking in the guts, or you have nothing to do (without downtime for other employees), pick to your health.

Answer 3

Often and densely, especially the workstation, reinstall is cheaper. If Google did not give an answer on the first page, then most likely the solution in the "correct" way will be more expensive than putting everything from scratch or butting with TP. Now, if a problem appears on a freshly rolled one, then it’s already to figure it out.

Answer 4

1. Make sure that the hardware is working properly. Often this is a hard disk (for example, an already broken, but not yet remapped sector that some process is trying to read; or problems in the file system - although this is not actually a hardware problem), or memory errors, or overheating. All of these are relatively easy to diagnose.
2. Make sure that you have the necessary rights and the availability of resources. These may be missing rights to some folder where the process is trying to write, or missing access to a network printer, etc. Diagnosis is a little more difficult, but still possible. Utilities filemon, regmon and something else will help.
3. Logs of the system, and the program itself (sometimes you may need to look for related ones). If they are being carried out. They may contain references to the sources of problems.
4. Experience, experience and experience again. Plus the ability to google and understand what to look for.
This is if you want to find the cause, and develop some measures to prevent the recurrence of the problem. If you need to fix it quickly, it's easier to roll back to a restore point or reinstall the problematic program or even the entire system. I prefer "lose a day, then fly in 5 minutes." But again, it depends on the severity and timing of elimination. If you have time, you can sort it out. If you have a backup computer - the most correct option is to put a reserve, and deal with the problem and find out the reasons. Then the deadlines are met, and the cause (possibly) will be found, and it will be possible to prevent repetitions.
PS There is still a nuance, for which the admin is paid - either for the number of processed applications, or vice versa, for the fact that there will be no more than a certain number of applications. And is there a record of work? If the accounting is kept and taken into account by the number of applications, it is unprofitable for a specialist to look for a problem, it is more convenient to quickly fix it and run to do the next one. And if they said "you can do nothing at all, just to make everything work" - then it's better to figure it out, set it up as it should and forget it. Freed up time can always be used in a better way than having to reinstall the program for one hundred and fifty.

Answer 5

The Lymar method is very effective - wait until it passes by itself, they get used to it, or they begin to consider it a feature.