Slashdot Mirror
RIM Releases Reason for Blackberry Outage
An anonymous reader writes "According to BBC News, RIM has announced that the cause of this week's network failure for the Blackberry wireless e-mail device was an insufficiently tested software upgrade. Blackberry said in a statement that the failure was trigged by 'the introduction of a new, non-critical system routine' designed to increase the system's e-mail holding space. The network disruption comes as RIM faces a formal probe by the US financial watchdog, the Securities and Exchange Commission, over its stock options."
More importantly, they apparently had no or a very bad backout plan.
It's quite likely the development group listed this as a risk, with a good backout plan, and upper management simply didn't want to pay for the cost of having a quick backout.
If that's the case, you can be pretty sure upper management WON'T take the blame.
You can tell this is a geek site. Bad software rollout, first post wants to blame the QA manager, second wants to blame "Upper Management." How about a little blame for the devs?
Blasphemer!
Clearly bugs originate with devs, the same way typos and spelling errors originate with authors. The occurrence of such errors is inevitable. The process as a whole is what is responsible for eliminating them. To the extent that the devs failed to contribute to that process then yes, they also deserve blame.
I am not sure if you are trying to be funny or insightful, probably you are aiming for a bit of both, however, while bugs in software (inevitably) are developers' fault, release of software with bugs into production system is always management fault. There must be a process in place to catch bugs before release for mission critical systems (isn't it one of them?) There must be a process in place for a quick rollback for such systems. There must be some form of backup. How about running both, new and old systems in parallel for a while with ability to switch to the old if the new one fails?
Whatever it is, the production problems are due to bad process, which is what management is supposed to control. They are not even responsible for coming up with the technicalities of the process, they are responsible for making sure that there is a sufficient process (sufficient in terms that it is agreed by all parties, DEVs, QAs, BAs, client that it is good enough.) They are responsible to make sure that the process is followed.
Over a year ago now in Toronto, ON, Canada, the Royal Bank of Canada had a similar problem, but of course with a bank it is much more dangerous it is lots of money of lots of people. Heads rolled at the management level only.
You can't handle the truth.
Yeah, they've got areas to tighten up their QA and patch processes, but on the whole they got it all back up and running faster than most enterprises get their email functioning after a worm.
Which is worse:
A) The fact one piece of software took down their environment.
B) Their failover plan didn't work.
C) All of the above.
D) None of the above.
Personally, I vote for "B". Face it, s**h happens. But when you plan for s**t happening and the plan doesn't work, that's a VERY bad thing.
This is my opinion. To make sure you don't steal it, it's covered by the DMCA.
And a bunch of suits will want the heads of the technicians responsible.
I feel for them, I really do.
A few years ago I put in a minor maintenance change that made headlines for my employer.
This is a natural result of the budgetary constraints we have to live with in the real world. Testing and certification is expensive, and the more complex the environment, the more expensive it gets. It is difficult to justify a full blown certification test for minor, routine maintenance, unless you are talking about health and safety systems. So a worst-case event occurred, RIM suffers some corporate embarrassment, some low-level techs will get yelled at, and possibly fired, and a bunch of people had to suffer crackberry withdrawal.
Nobody died. No planes crashed. No reactors melted down.
RIM will work up some new and improved testing standards, and tighten the screws on system maintenance so much that productivity will suffer, they may even spend a bunch of money on the equipment needed to do full-production-parallel certification testing. And then in a year or so cut the budget to upgrade the certification environment as 'needless expense', and come up with work-arounds to reduce the time it takes to get trivial changes and bugfixes rolled out.
I wish them luck. Especially to the poor sods who did the implementation.
At least when I did my 'headline-making-minor-maintenance' it only made the local papers for a couple of days.
http://visualizecommonsense.com/
How about a little blame for the devs?
Because that's not how change should happen in large/business critical applications.
What should happen is that the update is thoroughly tested, a change control request is raised and at the next change control meeting the change request is discussed.
The change request should include at the very least a benefit analysis (what's the benefit in making this change), risk analysis (what could happen if it goes wrong) and a rollback plan (what we do if it goes wrong). None of these should necessarily be vastly complicated - but if the risk analysis is "our entire network falls apart horribly" and the rollback plan is "er... we haven't got one. Suppose we'll have to go back to backups. We have tested those, haven't we?" then the change request should be denied.
As much as anything else, this process forces the person who's going to be making the change to think about what they're going to be doing in a clear way and make sure they've got a plan B. It also serves as a means to notify the management that a change is going to be taking place, and that a risk is attached to it.
And if a change is made but hasn't been approved through that process, then it's a disciplinary issue.
Of course, it's entirely possible that such a process was in place and someone did put a change through without approval. In which case, I don't envy their next job interview.... "Why did you leave your last job?"
I couldn't agree more. Yes, the developers should be responsible for their errors, but still, they're only human. Even the best dev makes a serious mistake from time to time. That's why it's essential to have good coders and good QA folks and good management for any project, especially one as large as the Blackberry network. Sometimes redundancy is a good thing.
RIM is not a regular company. They have specifically created a centralized system where the email for millions of people depend on the uptime of their two (?!?!) data centres. Delivering email is literally their business and uptime is a critical part of that. IMHO, a half an hour of system wide downtime is pushing RIM's luck.
Several hours of email downtime is "OKish" if you are talking about a medium sized company that only has a handful of servers and a few IT guys. This is not the same at all.
Prior to this, I never realized that the RIM system was THIS centralized. It's kind of concerning really. And I don't quite understand why so many US gov't users are allowed to route their email through a NOC in Canada (disclosure: I'm Canadian).
I am a dev and my motto is "all software engineers are liars and idiots" and I include myself in this. If you want to know how something is supposed to work in theory, ask the dev. If you want to know the actual behavior, ask QA.
Avoid Missing Ball for High Score
You can't expect programmers to do perfect work, even with unit testing and all the other basic amenities of software development. It requires QA, and that is something sorely lacking in contemprary software product. From the smallest OSX widget to MS Vista,Testing Matters.
RS
Shoes for Industry. Shoes for the Dead.
As a QA guy, I can't tell you how many times I've been told, on a Monday, "Do whatever is required to make sure this software is stable, as long as you release it on Friday."
:-).
We're lucky we can get through a single pass of functionality testing; forget about load/stress/performance/long-term stability. We're lucky we have a test environment composed of hardware retired from production, because it was deemed insufficient to meet the needs of the production environment.
True story: I was supposed to be testing a product that interfaced with an IP videoconferencing bridge. Except we had no such bridge in our environment, and no budget to purchase one. No one in management thought this was absurd until I took a cardboard box and wrote "Video Bridge" on it, along with little holes labeled eth0, eth1, DS1, etc. (much like the famous P-p-p-powerbook). I complained to the VP of Engineering that our tests were blocked because I couldn't get the video bridge to come up on our lab network. When I showed him the "box," he got the point.
In my experience, customers are more interested in getting new features ASAP than they are in reliability, which is why so many organizations put a premium on rolling out new features quickly. When was the last time anyone worked on a release with no new features outside of performance and stability improvements?
How many people here have checked in buggy code that neither management nor QA knew was buggy? (crickets)
How many people here have been on projects where management shoved the code out the door despite major bugs that they knew about? (thunderous applause)
How many people here have tried to get time on The Schedule to do something The Right Way, only to be told by management to do it half-assed, because that's all there's time/resources for? (applause, hooting)
There you go.