Sonobe Lab. > > > Lc| Japanese|Ger|Fre|Spa|Por|Ita
** Sonobe Lab for 22nd C!**
How to investigate system troubles smartly How to investigate system troubles smartly  
* This page is made by "machine translation" program from Japanese to English.

Although work called trouble investigation and debugging is thrilling so that it may be whether he became a noted detective or a noted detective if it succeeds, it is that of convex potatoes running into difficulties.

If the following is met and troubled by . trouble which was invented from my experience and which is advice for trouble investigation, this check list should be used for it.

that pleasant-sensation . which runs closely after a criminal (conviction insect ?) who is going to run into a labyrinth, or waylays, is made to call existence ガチャリ, and arrests いわさず handcuffs -- in order to taste that, it is reasonable and a device which is not caught by blind spot is effective

This page will be the highest if you are useful to solving quickly a trouble which will encountere from now on.

[ ] Wasn't an already known trouble encountered with?
Or [ that it is where if written by paper, Web, or electronic manual about a trouble ]

[ ] Aren't there those who have encountered the trouble that otherwise it is the same in fact?
There are fairly many people who are silent even if it encounters a trouble.
-> I will ask a man of . circumference who will search in and outside the company with Web.
I have the same thing as a surrounding man done, and will have it shown whether a trouble comes out.

[ ] Isn't it normal in fact?
There is often misunderstanding of specification.
What kind of specification as -> specification then is it considered? I will think on a desk.
It checks and needs.
The developing agency may have changed specification.
Specification has changed before that it is unexpected.

[ ] Where was that was normal recently when?

[ ] When normal recently, what did themselves, others, or a computer change after that?
-> I will imagine how unusual hardness and unusual software, and a network influenced a trouble.

[ ] Didn't it happen by upgrade?
[ ] How was it in an old version?
[ ] Isn't it the problem of non-compatibility in a high version?
[ ] Didn't any set points increase?
[ ] Has trouble correction hit perfectly?

[ ] It carries out a reappearance test by large environmental condition including conditions from which a trouble arises and needs.
-> . which will look for the boundary point [ makes the environmental condition (execution root) smallest, and / a trouble ] no longer coming out of -- there is an insect on the boundary point
A "trouble mini model" and I have named such minimum environment that makes small with a grinding sound to the boundary point, and a trouble reproduces.

If a trouble mini model can be made, when . trouble investigation which can already solve く障害 will not progress at all and will be irritated by analyzing conditions of a small number which constitutes it, I will do making a trouble mini model hard.
Although it is however afflicted by trouble when there is little repeatability ....
I will write out and look at an environmental condition and happening /起こらない.
In order to make an environmental condition small, I will look for whether there is any function which can be bypassed.
I will write out an idea "whether OO will be bypassed."

[ ] It carries out a reappearance test by such small environmental condition and needs that conditions from which a trouble arises are not included.
-> . which will look for whether a trouble comes out by enlarging the environmental condition (execution root) more, and that boundary point which does not come out -- there is an insect on the boundary point
Such greatest environment that a trouble does not reproduce is named a "normal Max model."
I will write out and look at an environmental condition and happening /起こらない.
In order to enlarge an environmental condition, I will make it pass along a part of environment where a trouble is encountered.
I will write out an idea "whether it will also let OO pass."

[ ] A method "the minimax clip 打ち method (mini-max squashing method)" of shutting up and crushing an insect of a between by an above-mentioned trouble mini model and an above-mentioned normal Max model is practiced by itself for years, and there is confidence in the outstanding effect.

Although it differs from "a trouble OFF division" generally said, a trouble OFF division may also be doing the thing same in fact as this depending on the case.

At any rate, when a cause of a trouble is not found logically and is puzzled, if a trouble mini model and a normal Max model are made hard, it is glad that there is a path which a cause understands.

inside which is pursuing a normal Max model -- "-- it is -- the present purpose cannot be achieved if it carries out like this -- " -- a "recess way" which can be employed even if a trouble is not solved can be discovered in many cases
Is a recess way already found? Otherwise, I will search from a normal Max model.

[ ] When a bug is where or it says that between the present trouble mini model and normal Max models will be sampled and tried, since it is not necessarily a speedy, it needs attacking from an edge gradually carefully.
For example, if pass with a route of 1024 is attacked from an edge, supposing there are all possibilities, an average of 512 trial will be needed.
on the other hand, . which may be good to adopt "dichotomy (binary division method)" -- a method of this dichotomy that I considered is as follows

if butterfly middle (と -- a hit considered) of a point acting as a trouble and a point not becoming is tried and it becomes a trouble, as it says that a center is tried further, it will be the method of the point and a point acting as [ if a center is tried further and it does not become a trouble ] the point and a trouble of a point acting as a trouble of contracting the section with a half, a half, and a half
Then, even if there is a route of 1024, it is attached . good for the boundary point by a maximum of ten trial. . which can be referred to as very efficient (the number of log2 routes)

for example, . which somewhere in some files have the character of a bad code, and garbled characters of a file generated by editor when saying in an example of a trouble of one day -- [ there ] . which narrowed the section for whether it will be transformed if even the section to the how many lines of a file is extracted -- which is such a condition -- [ 500:NG, 250:OK, 125:NG, 190:NG, 200:OK, 195:NG, 197:NG, and 199:NG ] When a bad code was in the 199th line, becoming clear and it were removed and solved.

If supplemented, time and effort may not require a direction attacked from . edge which cannot say dichotomy with the shortest time to a flume difference which considers cost which creates test environment.
probably, it will be good to investigate a place which can be easily examined by return by part for an inner part of . company which may be one with it better [ to leave at the end ] by dichotomy inside of it, since making a trouble application and investigating to a telephone carrier will require cost rather than an end reason for an inner part of a shrine, if it says in an example of line failure investigation

Moreover, when investigating where [ of a program ] a variable has broken, it is thrilling not to be from an edge, either, to devise a snapshot at many points simultaneously also by dichotomy, and to make it become clear by one shot.

[ ] If program such a trouble phenomenon is said conversely, be easy to realize making it what kind of structure, or design on a desk.
-> . which may actually have become so -- if possible, tips tend to cause the trouble event by one easy change

[ ] others -- a trouble and a doubtful phenomenon which have happened on the outskirts -- what There is relation.
It has sometimes often happened by the same cause.
-> I will consider whether it realizes by saying "Carry out programming which starts both simultaneously."

[ ] . which may have induced a new trouble by mistake when a trouble is corrected -- there are very many these patterns
a trouble which themselves and 1 corrected recently [ -> ] -- where -- だっけ Even if relevance is unknown, it will check a corrected portion finely for the time being.

[ ] Is there one more same thing and any mixed up?
-> I will consider whether there was nothing that is a little different by the same program name.
Supposing it is not confusion, I will experiment also in the similar thing.

[ ] It may be shortage of knowledge.
Isn't there any applicable part in a related manual or an electronic document?
-> it warns against . which will read and will look for a hint, however repenting of reading in large quantities aimlessly just and having spent many hours too much.

[ ] It may be a set-point mistake.
. which will remember a set point of which -> relation is done -- I will have those who set up correctly and are moving teach a set point

[ ] Didn't unstable hardware or an unstable circuit induce a bug of software?
Can't it try on hardware or a circuit which were stabilized more?
Or [ if it is the telephone line, / it will make it another circuit ]
There is also a way of making a trouble reproducing with all its might, and investigating, using purposely the hardware unstable on the contrary and circuit unstable on the contrary (in the low case [ Repeatability especially ]).
It is also good for it in a world to reproduce . trouble in which a person who excelled in if operation is busy, stirring up a trouble by Tati who wants to use various functions is (I am also kana so) to have such a man use.

[ ] Isn't it the trouble which dangerous timing induced?
Didn't it break in parallel processing? It investigates whether more than two (two persons) were moving simultaneously.

[ ] It examines peculiar trace of the trouble and needs.
How depending on which something breaks, an error message, timing which comes out are made [ collect the various features of a trouble, individuality a strange point, and time of "smelling" that does not seem to be accidental somewhere, and ] a key of criminal specification and needed.

[ ] It investigates [ as it is made visible and can be heard ] and needs.
If it is its program, even if it will carry out the snapshot of the form in receipts and payments of important information and he will think that it is like, . that it is natural, it investigates whether it is satisfactory in that.
That out of which noises, such as a modem, come will make a noise with a set point.

[ ] I will write an information flow by box and arrow.
Which box carries out the one-piece one-piece check mark of whether it is クロ and needs whether it became unusual and which box from where at シロ.
. which must not have the box which is not written from the beginning noting that it is シロ -- somewhere -- シロ -- that is right and there is クロ い箱 -- because make a mistake in being alike and there is nothing -- シロ -- you have to probe just a box with that right perfectly

[ ] Also while doing trouble investigation by one oneself or a small number of people, I will carry out the log of the investigation process to memo paper, a note, or an electronic file.
It records distinguishing a becoming-clear fact etc. as a result of a hypothesis and execution.
A new leading hypothesis comes out with ま見直してみる which came, or a becoming-clear fact of being buried becomes clear.
. sharable as trouble investigation know-how if it collects on an electronic file and will carry on Web (. which is not an object for individuals although there is saying in a tool for support centers (CTI), such as Primus, in a world and record and a retrieval can be made reasonable)

[ ] I will consult with someone with information on which many hours were not spent recklessly but which collected them alone until now. .

- Copyright ©, Masayuki Sonobe 2002-2024.
- Updated 5359 days ago: 2009. 7. 2 Thu 19:20
- Accesses to the Page since 2002.09.09 00,005,403
- Accesses to Sonobe's since 2002.07.01 05,076,096