Abstract is a midnight emergency calls, I hurriedly rushed to the research room GTAC deal with urgent problems, a door, only to find a few before the old man: "See you come to my mind." This sentence is a time to let me myself full of energy!
20 million lines of code, do a set not?
"You carry the shoulders of 20 million lines of code, do a set not?" This is the most heard a word when I first entered Huawei.
In September 2008 a new employee training was finished, I did not return to Xi'an, but directly fly to Shanghai to undertake product. It first entered the Shanghai team, they received a warm reception.
"This Xi'an How many people came to be our module?" PL Shanghai X modules ask me.
I Leng Leng pointed to himself: "Just me."
"What, you alone? We have a group of 200,000 lines of code, and quickly feedback plus people."
In fact, I just left the campus of the 20 million lines of code and there is not much concept, but to see his reaction, he felt good. I quickly found the product to undertake projects in other groups Xi'an PL, but was comforted, "did not think so serious, you piece of business I know of a number, I'll give you the code logic analysis see .X module is relatively simple, Y basic module no problem may temporarily not learn, so ...... so ...... as long as you focus on these few modules to get on the line. "
Plus people is not possible, so my plans have changed accordingly. Entry first two months, my daily task is to read the code, before work half an hour to explain the whole group. Other new employees over the same period in the month of entry have begun to deal with the problem single and development needs. The first three months of interim reply, presented on other people's films are "dealt with XX a questionnaire, developed demand XK code," and my film is the module code Features.
After the end of the study period, 200,000 lines of code might soon be reflected. In order for me to quickly familiar with the business, the project team to arrange the interface people live the most sophisticated work to me, he is responsible for all online issues, problems and problem locating a few high-risk laboratory module single modification. Soon, I was drowning in a sea of telephone and e-mail, the bruised and battered.
"The problem is not positioning? Messages are sent for several hours, the environment is not retained."
"The problem I should be the analysis of your module, quick look, I did not work before the conclusion to turn up the order."
"How is it, how you name a single issue but not less, faster version could not pass the point."
……
Yes, I became the "focus" of each group, but also began to become anxious, though only back to the apartment every morning, still can not stop the DI (questions density) channeling, which is what to do next?
Soon, mentor and PL noticed my dilemma, lent a helping hand. Look phenomenon, to find a cause, and I analyze the situation, one by one analysis of the matter at hand, to teach four-quadrant work experience, the matters lined up in front of the four-quadrant priority in accordance with the law, break 11 again, slowly persist for some time after I finally became calm them.
It is to sleep soundly at night
After the positive, I soon encountered the first online issue, and experienced a profound process of bug fixes. Vaguely remember it was Sunday, the two in the morning, when the phone rang I felt like it was just lie down and wake up.
"I was in the test line support manager, a new version upgrade error, we need to be solved at once!"
"Yes, what? Do these steps first attempt at a recovery operation, and then collect under the log, I'll go to the company." Is a problem of the existing network, we had a look of confused me instantly awake, it can not tolerate any careless. Quickly dressed and went straight to a taxi company.
Fortunately, the problem quickly locate previously existing network server network card failure occurred, the customer opened the server, two network cards to pull out wipe Goldfinger, insert the time slot exchange position, leading to the PCI bus card numbers change. In order to prevent the unauthorized installation of other customer card, causing compatibility problems, a new version of the code to do a mandatory check, but for such a replacement card position of the scene, but did not take into account.
"Who designed this function, superfluous! Old version no problem, this is a fatal bug, I asked to be back!" Although I am not a function of the development, but to hear the other end of the phone harsh wording, but also feel like a made a big mistake , did not dare say a word. This experience, let me follow-up for a long time, received a phone-line problem is very tense.
Internet problems caused controversy in the past have not, modify the online issue of the task fell to my head, did not think that this amendment is not smooth. Modify the code was soon over, but encountered a problem verifying. Because the stock is not much older model server a few years ago and has stopped shipping the three old model of the server, only one laboratory, and the other two types of servers is no way to verify how to do?
"Code is to determine these information, you can simulate piling, before I have been so measured," under the guidance of older employees, I quickly completed the piling test, but somehow my heart is not practical.
The results in the pre-verification link before the turn of the internal test version, colleagues, brothers and project team to find me, they run a server installed after the new version is not up. I heart "Ko Deng," look, those two will not be no verification of the server, right? Sure enough, after the real machine analysis, I found that piling simulation and real hardware there are still differences.
For this revision introduced, PL specially come to counseling: "This is a major lack of experience you do not mind too much, but we have to think about it, difficulties, is not did the best.." Revise again, there is a type of server is not found, feeling really no other way.
Early in the morning after I had again to help mentor and PL, hours, PL came to me: "I have made a phone call to several neighboring departments, there are several servers may be we're looking for, I'll take you confirm under.
After another couple of hours, we finally found a corner of a laboratory of a server dusty. Dust ourselves off, a look, boy, is not that what we're looking for it! Looking for power, wiring, power-on, install version ...... version of the software to see a smooth start, was suspended for a long time and finally landed a big stone.
"Well, then we switched the order of the scenes at the card cover." However Shetenglebantian, pull down the card or not. The original design of this model on the server hardware also made foolproof, unload the card requires a dedicated gadget.
Half an hour later, the card was finally pulled down, PL hand was accidentally struck a hole and bloody, he was quite not care: "In order to sleep soundly at night, these injuries, the value!"
Later, whenever the "word of mouth build quality, build quality cultural church" See, I want to say our pursuit of quality is really very simple, just to sleep soundly at night.
No not locate the problem
"Hello, sir ......" just left the company after work, I received a foreigner phone. Ears coupled with skilled "sorry" "pardon", it finally became known, turned out to be talked to before in India espace line guy, immediately go to customers operating room, there are two steps less clear.
Never with foreigners over the phone, I know what to say, red in the face, mouth several times to Zhang, but is not open to the word of mouth, it is unable to speak.
The other is still time to time "hello? Hello?" I do not think that online. Hey, I usually help a brother GTAC communication, which under no one to help, which can do ...... No matter, how his pipe syntax, suddenly, a "yes" jumped out of the mouth, slowly, a word, a word to pop out, though stammered, but at least you can communicate in English the.
I said on the phone a pass, and after repeated confirm that they understand what I mean, just put down the phone. A look nervous palms are sweating. Fortunately, the exchange finally finished, I felt a refreshing lot.
In this way, entry after two or three years, everything gradually on the right track, no matter what the problem is encountered, it seems I can calmly deal with the.
But two or three non-targeted Linux systems from the network now appears linked to death problem, everybody has been shrouded in dark clouds overhead. Because Linux using version purchased a few years ago, has not been upgraded, dimensional measurement function is relatively weak. And we as a business software team and do not have experience in locating this difficult problem, seek help after the company's Linux team, still can not be located, only a Linux system responses to old problems.
I did not expect, and soon, in a big game point T and the emergence of this problem. No way, we can only help the OS, hardware-related personnel again, quickly set up a research team. Due to the lack of logs, software and hardware everyone from a variety of angles bold speculation, then fault injection testing in the laboratory, lasted a month later, the lab did not even reproduce the problem, let alone locate, everyone was depressed . Fortunately, the new version of the software have been combined into a self-healing function is automatically reset when linked to death, may be affecting the problem to a minimum. Line also received a self-healing program. For the first time focus on research so unwilling to end in failure.
Out to mix, sooner or later have to repay. Research problem will never be absent, but it comes late. After six months, along with China, the problems linked to death Linux kicked off the second research. Because it is out of the question at night, not enough time to first-line treatment, we request line reservation environment, immediately coordinated the company's Linux and hardware specialists immediately to the scene to locate a business trip.
"This caught the first scene, finally able to locate up." I thought. Unfortunately, he did not return the good news from the front line, just further confirmation, indeed Linux system linked to death, the reason was not known. Suddenly, he reached an impasse.
But research can not therefore stagnation. We stop again, to continue to analyze the log, look at the code, analyze these questions to find common ground, and soon found that these devices are the issues involved in running around when there is a problem almost a year. Could long-running relationship with the board when? The cumulative effect of the failure mode? He quickly adjusted the research direction. Soon, they find a case on the Linux kernel memory leak, after calculation, our board just a year or Linux system will run out of some of the key memory, the system appears to hang dead. Really beyond the pale in hunting, must come not a waste of effort.
I participated in the follow-up and even led many incidents time-consuming, difficult problems arduous research, Linux system linked to death research in these issues is not the most influential and most urgent, but let me most benefit. Having gone through this research, to face any difficult problems, my mind has a belief: In Division I, never not handle things, and I never locate the problem!
2014 to 2017, due to the adjustment of work, I fought in product development, temporarily left the problem of online processing. 2018 I once again return.
Is a midnight emergency calls, I hurriedly rushed to the research room GTAC deal with urgent problems, a door, only to find a few before the old man: "See you come to my mind." This sentence is a time to let me myself full of energy!
Source: Huawei published within the "Huawei people," Do not reprint