Slashdot Mirror
Debugging The Spirit Rover
icebike writes "eeTimes has a story on how the Mars Rover was essentially reprogrammed from millions of miles away. 'How do you diagnose an embedded system that has rendered itself unobservable? That was the riddle a Jet Propulsion Laboratory team had to solve when the Mars rover Spirit capped a successful landing on the Martian surface with a sequence of stunning images and then, darkness.' The outcome strikes me as an extremely Lucky Hack, and the rover could have just as likely been lost forever. Are there lessons here that we can use here on the third rock for recovery of our messed up machines which we manage from afar via ssh?"
Are there lessons here that we can use here on the third rock for recovery of our messed up machines which we manage from afar via ssh?
As a former co-worker (hi, jwalker!) used to say when people tried to draw ridiculous analogies, "It's exactly like that...only different."
A programmer is a machine for converting coffee into code.
Man, I have a hard enough time debugging programs running on my local machine.
I dont think i want to learn too much from this as the solution was the equivalent of rm -rf... On a side note i wonder when the 40 min ssh delay jokes will begin again
drunk chemists
at least it wasn't a blue screen?
man rover?
The Human Cow - bringing you scrumtrelescence since 1995
I don't get it, couldn't NASA afford the on-site warranty?
Wow, I didn't expect the rover had 128MiB of RAM, or 256MiB of flash. Funny to think they had to run chkdsk from so far away :)
That's the thing that amaze me. Any technology having to do with space seem that much more advanced.
Here on earth we can't even build cars that require no maintainance and last more than 10 years.
I hope they use SSH or something .. who's to say a future mission ..some hax0r doesnt grab control of a space probe and have it send goatse.cx pics back??
.. after all the probe communicates using known frequencies. There may be probs picking up the return signal without an expensive antenna i suppose. But then again maybe some hax0r can build one cheaply and or do what captin midnight did ( www.signaltonoise.net/library/captmidn.htm ).
All it takes is a transmitter out in the middle of nowhere africa or some island
I wouldnt worry about signal jamming though as that will probably be discovered easily.
The Martians are pissed that the repair labor was outsourced to Earth.
Table-ized A.I.
If it was the hardware that got fried and they miraculously fixed that, I would understand but this was just a software glitch.
I routinely reboot and reprogram machines in our data-center that is 2000 miles away from me.
As long as all hardware components are working and there is connectivity to the machine, it doesn't matter whether the machine is a few miles away or a million miles away.
Sounds like NASA forgot to empty the rover's recycle bin. =)
Steal This Sig
...would have been to have "fixed" the problem before the hardware left earth. This "bug" (or more accurately, known limitation of the filesystem) should have been discovered here on earth if the rover had been properly tested.
The only real bug was the inability of the system to properly handle running out of file entries (or more specifically, consuming too much RAM as the number of file entries increased). However the software should have never have stressed the filesystem to that degree in the first place.
Dan East
Better known as 318230.
Granted mainstream media have to keep their coverage dumbed down if Joe Public are going to read it. But what really bugs me is the lack of follow-up. We hear about poorly understood events as they are unfolding, then never heard about them later when they are completely understood.
A recent example is the gangway between ship and shore at the QM2's drydock. It collapsed killing lots of people, an investigation was launched. Why did it collapse? At the time it wasn't known. I'm sure it's known now, but there's been absolutely no followup.
This article about the rover is great not so much because of the level of detail but because it reports on an event with the benefit of hindsight.
Slashdot monitor for your Mozilla sidebar or Active Desktop.
What filesystem is used? Is wear leveling being used? The directory structure is apparently stored in RAM during the day (why else would it use so much RAM?), that is a good thing for reducing wear on the flash system. But what's the number of writes on the flash chips? When will that number be reached?
"It's too bad that stupidity isn't painful." - Anton LaVey
'How do you diagnose an embedded system that has rendered itself unobservable?'
The way you do this is by having an exact duplicate of the remote system so you can set up a test with conditions as close to those under which the remote system is currently operating. You can then do a series of carefully controlled test solutions to determine the optimum prior to trying it on the "live" system.
This is the way I set up all my production systems and, barring catastrophic hardware failure (self-immolating disks and a router which just folded when its power supply burped) I've had perfect uptime.
(well, ok.. there was that one time, late at night, when I typed "reboot" in the wrong window.. but that happens...)
I have something in common with Stephen Hawking...
If you RTFA you will realize that I'm not lying in the least when I say that, effectively, they ran out of flash-based "disk" space! They forgot to delete old files when updating the programs in the flash memory (which is mounted like a filesystem, or hard disk), and the OS was failing because it wanted to use that space. So it rebooted, and still had insufficient disk space, and rebooted again . . . lather rinse repeat. There was no signal because it was stuck in a reboot loop because they ran out of disk. Wow.
:) ), then used low-level (direct access) flash utilities to remove the old files. Reboot, mount, disk check / corruption repair, voila it works again.
They fixed it by telling it to boot without using the flash (safe mode
We have a big 1TB NetApps server where I work, and we have so much disk space that people get lazy and don't delete files or archive old projects, then they get really confused when jobs fail, not thinking disk space until checking everything else first. But it happens, and it's usually surprisingly hard to debug (they check a lot of other things first, sometimes even upgrading tool versions!). It's really kinda funny, in an expensive and mildly embarassing way that the Spirit had the same problem.
everything in moderation
"The outcome strikes me as an extremely Lucky Hack..."
The outcome does not strike me as a "Lucky Hack." They made the system flexible, that flexibility got them into some trouble, and it's also what got them out of it. Anyone else agree?
Yeah, that was HAL's excuse too.
Seriously, hats off to all the JPL programmers. Proving to the Martians that there is indeed intelligent life on Earth, very intelligent.
My pet peeve when I'm doing remote troubleshooting is 'ifconfig eth0 down'...oops. At least NASA is smarter than that.
Peter.
You know what I hate? Wait, what do you like? I hate that!
MoFscker
Another factor in this is the safety of the flash ram. It is rad-hardened and built with tons of extra error correction which again, requires years of testing and special design considerations. And is extremely expensive.
-
Your post is the only thing that strikes me as a "Lucky Hack" here. They included the ability in the design to remotely disable booting from flash and upload new boot images, in what way is that a "hack"? All this is just foresight in design to include as many possible recovery modes as they could.
Basically, they rebooted from a recovery image (sent via radio) and then proceeded to do low-level fixes on Flash memory and they a chkdisk. If I do something similar via recovery disk or CD, I don't get a lot of people telling me that it was a "Lucky Hack" that I could boot off of CD!!!
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Great article! This is just the sort of thing that has always impressed me about NASA and the JPL. Just when mere mortals might give it up and walk away, they figure out the problem. I can only imagine how wild the party must have been after they fixed Spirit, the scientists and engineers I've worked with in the pass could really put away the booze.
Seriously though, the key lessons to take away from this are.
1) Gather all of the clues you can.
2) Take those clues and build a model.
With luck and care, the model should get you closer to what may have gone wrong. And in this case it apparently did just that. Now that's geek cool!
BTW, I know that generally you want to prevent this sort of thing from happening. But in reality most software ships with bugs and launch windows to Mars are non-negotiable.
To the making of books there is no end, so let's get started
There also needs to be a way to load bootstrap code remotely. For instance, having a TCP/IP enabled BIOS be able to run TFTP or some other protocol to load a netboot floppy image. Then you could give it a LILO command instructing it where to find a boot image, preferably one on a server in the same hosting center.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
Operating System not found. Press any key to continue.
Damn! Left the floppy in!
What surprises me is that they don't have a 'twin' of the rover's computer system set up on earth. When commands are run on the rover, the same commands could be run on the computer system on earth. Then, if the rover's software, fails (as it did), the software on earth would (theoretically) fail in a similar way, and be MUCH easier to debug. Of course, the systems wouldn't be identical (without building an entire duplicate and expensive rover), and the data gatehred wouldn't be identical, but if the twin was carefully planned and fed dummy data that aproximately mirrored that data the rover was gathering. For example, the twin could be fed dummy pictures about as often as the rover took a real picture.
From the article "[The] transmission that uploaded the utility was a partial failure: Only one of the utility program's two parts was received successfully. The second part was not received, and so in accordance with the communications protocol it was scheduled for retransmission on sol 19." NASA could have simulated a half failed transfer on the twin copmuter on earth, and then watched carefully using traditional debugging tools to make sure the failed transmission didn't cause a software failure (which it did).
Again, from the article "The data management team's calculations had not made any provision for leftover directories from a previous load still sitting in the flash file system." However, if they had a twin computer system to watch, they would have seen that the failure occur on earth as it did in space. Debugging a system you can hook a serial debugger to is bound to much easier than debugging a system a million miles away.
Stupid like a fox!
To me, if this were a Unix-like system, it sounds like they ran out of inodes. Running out of inodes is very different than running out of disk space.
If you think runing out of disk space can be hard to trouble shoot, try running out of inodes.
Wrong wrong wrong, as I'm sure someone else will post. He spins a good yarn but he's just a machine room flunky and hasn't RTFA himself.
One simple rule for its versus it's
I had pretty much the same post - the originator of the story confuses luck with skill, a mistake a find very annoying and committed all too frequently. I'll fully admit when I've been lucky, but I also went recognition for foresight when I've had some! NASA deserves at least that much respect.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Duh. That's what they have been keeping a secret. They have a DB9 serial link strung from here to the landing site. It's not as cool as you all make it out to be.
First wxWindows, now Vx-works?
Before doing something risky, type this:
:-)
:-)
sleep 600 && reboot &
Now if your risky maneuver makes the ssh session unusable, just wait 5 minutes for the machine to reboot.
This is great for fiddling with firewalls by remote control... through the firewall.
Oh... You say you're not using a POSIX-like system? That's not supported. Sorry.
They could have set it up out in my backyard to take pictures of the piles of crap and rocks out there and if they wanted to simulate the solar radiation, they could have my girlfriend give it one of her famous looks... cause those are leathal enough to burn a hole in your soul.
-SF
That must have been some feat to get the arm on the rover to press Ctrl, Alt and Delete at the same time!
What really surprises me is that NASA did not verify the software. Software verification is essentially mathematically proving the software. It is tedious and expensive but we are talking about NASA and the Mars. Infact even beloved MS formally verifies device drivers before use ( believe it or not !!) If the original program was correct they wouldnt have to reupload it and the entire problem ...gone.
Today we salute YOU, Mr. Super Wizard Windows Reinstaller.
Only YOU can fully appreciate the difficulty of running a format c: command, while swilling a room temperature can of Red Bull.
"Hey this stuff is hard now!"
While NASA is too preoccupied with things like farway rovers, you take your vocational tech school fueled arrogance directly to the place where it will make the absolute least possible impact: A Slashdot discussion thread.
"Loggin' on now!"
Your unique eye for obviousness allows you to sling turds of obtuseness every which way, and then brag about how you were RIGHT as soon as one of your pronouncements hit true - regardless of how many times you were wrong before.
"See I told you sooooooo!!"
And if some idiot rocket scientist has the unmitigated gall to not bow down to your obvious Geniusdom, you unleash your fury down upon him with all the tenacity and mercilessness of a rabid pit bull with a tender buttock locked in its jaws.
"Total anonymity!"
So keep clicking away, oh Marauder of the Mousepad. Because when the results you so desire finally come about years from now, you can say it was because YOU demanded it."
"How come they haven't fired that dumbass head of NASA yet yet?"
(Bud Light Beer, Anheuser Busch, St. Louis Missouri.)
Using the low- level commands, about a thousand files and their directories -- the leftovers from the initial launch load -- were removed.
I think that means they deleted the useless stuff they wanted to delete anyways but didn't get to delete before the crash. I also remember news about science data from before the crash that was received after they got the rover working again.
As for how critical it is, well yeah, it seems the rover didn't need the contents of the flash file system. The operating system and other software was in the same flash memory but I assume that any sane designer would put in some hardware write protect interlock that's not easy to defeat accidentally.
You realize that missions to Mars can only be launched once every two years, right? If they miss their launch window, they've got to wait two years before they can launch again.
You also realize that NASA did do a test mission, right? They built a test rover and put it out in a desert somewhere. They used the mission to test the hardware, test the software, and to help train the team.
You realize that the onboard computer is basically the same one as used on the Mars Pathfinder lander, right? Same CPU, same amount of RAM, even the same OS. I wouldn't be surprised if they used the same (or similar) circuit diagrams for certain things.
The point is to use well known and well tested hardware. The whole point of Mars Pathfinder was to develop a system whose design could be re-used for other Mars landers and rovers.
Lastly, what exactly are you going to do with greater flash capacity? The point of having any flash memory on the rovers at all is not for long term storage, but rather just to hold onto data until it can be transmitted to Earth, after which it gets deleted.
Despite what some idiot posted a few posts up, they did NOT run out of room on the flash drive. Rather, the problem is more akin to running out of i-nodes. Mounting the flash filesystem, reading all its metadata and whatnot, took up more RAM than was allocated for it, due to the high number of files it had to deal with (most of which were accumulated on the way to Mars, and were going to be deleted).
is because when the batteries got drained the os went into a stable "safe mode" state. If they made a long lasting powersupply this project was doomed(.f) and they never found out what the real problem was.
Actually, they used VxWorks because it was the same OS used for the lander on the Mars Pathfinder mission. Since they were using the same CPU and same basic computer design as the Mars Pathfinder lander, they probably figured, "Why not use the same OS?"
I'm serious. http://physics.nist.gov/cuu/Units/binary.html for all the groovy details. If anything, it's a move away from the hd manufacturers lingo.
Here's what happened according to the article. They launched the ship with an OS image in flash, and soon realized that they needed to update it. So shortly after launch they sent another complete OS image. They knew they'd have to delete the first image, but they didn't do it right away. At that point there was plenty of room in the flash memory so having two OS images was not a problem.
After a few days on Mars, they were starting to fill up the flash, so they planned to go ahead and delete the old launch OS image, its directories and files. This is a complicated process so they uploaded a special program to do it on Sol 15. And apparently they informed the rest of the team that the memory would be free and available after that point, so the rest of the team made plans to start filling it up with pictures.
However, the upload on sol 15 failed, and was rescheduled for sol 19. Now, here's the big mistake (which the article glosses over): They forgot to tell the rest of the team that all that memory wasn't going to be freed up as planned, not for a few more days. So instead, Spirit is moving around now, taking lots of pictures, storing them in flash, and all the people involved with that think they have plenty of room. Little do they know that they are running out of flash space. Finally, the morning of Sol 19, shortly before the memory cleaning program was going to be sent down, it happened. The flash memory was exhausted. This triggered a sequence of events which put the craft into a failure loop.
The big problem here, then, was the failure on the part of the group which was supposed to clean out the launch OS image to tell the rest of the team that it wasn't going to happen as scheduled, so the memory wasn't going to be available. It wasn't really Murphy's Law, but rather a failure to communicate among the team. This is an institutional problem which will hopefully be fixed.
It was the inability to build the RAM-based directory structure of the files in the Flash memory.
Why couldn't they build the directory structure? They had too many files, the size of the files doesn't matter here, only the number of files.
In other words, they ran out of RAM, not Flash.
Exercise left for the readers: Why can a Unix file system that is out of inodes have much less than 100% disk usage and still not be able to create a file?
If you're really worried about your remote server being unreachable, here's what I would suggest doing:
Have a hardware watchdog. If the machine is lost or confused, it reboots itself.
Have it come up in a known state, fire off a few broadcast packets to the sysadmins, and run sshd but basically nothing else. Stay there for a minute or so.
If nobody's tried to log in and halt the boot process, carry on booting. With luck the problem was transient. Worst case the problem still exists, you reboot, and the admins get another chance to log in.
From the description of how they got Spirit back, it looks like this is exactly how it was set up.
Who'da thunk it!!
455fe10422ca29c4933f95052b792ab2
WindRiver may give JPL large discounts, but I doubt that's the only reason VxWorks is running on the MERs.
Years ago, when JPL was designing the Mars Pathfinder mission, they asked Wind River to do an "affordable" port of VxWorks to the RAD6000 (a radiation-hardened RS6000), and they agreed. Since the computers on the two MERs are very similar to the computer on the Mars Pathfinder lander, it makes sense that they'd use the same OS that they used on the MPF lander.
I would think the fact that JPL knows VxWorks very well by now would be a major factor in deciding to use VxWorks for the MERs.
The JPL is a pretty viral license. It forces you to spread their space probes from your planet to all your customer's planets. This is un-solar systematic! What's next? Calling GNUpiter Jupiter instead?
One word: outsourcing.
When I worked at JPL, every 6 months to a year there'd be talks of layoffs because the headcount was too high; people would leave and return to the same projects as contractors, then get a higher hourly wage for doing the same work with less accountability.
The whole reason for that lost probe (feet vs meters, anyone?) was because of a political squabble between two teams (one JPL-internal, one outside contractors as I recall) who simply failed to cooperate productively. The whole management structure inside that world is screwed. People's project leads are not the same as their section/department leads, so the reporting chain is a mes{h,s}. Time and energy is wasted in contract(or) management, all in the name of "reduced costs" even though having all the work done in-house would eliminate a full layer or two of mid-level management waste.
NASA/JPL are totally hamstrung by beancounters who think they're saving the public's money, but truly can't see the big picture, missing the forest for the trees. (Either that, or they *do* see the big picture, and are busily lining their own pockets with the excess that gets tossed around thru all the churn.)
-- *My* journal is more interesting than *yours*...
Could this have not been said more succinctly with a simple quote? Namely:
"What we have here, is failure to communicate."
Mal-2
How is the Riemann zeta function like Trump rallies? Both have an endless number of trivial zeros.
"The irony of it was that the operating system was doing exactly what we'd told it to do"
Funny, that's how it was explained to me by my computer science teacher my freshman year in high school. He said, "The problem with computers is that they do exactly what we tell them to."
I belong to the ______ generation.
considering the distance i'd say a while, couple hours doesn't make much diffrence when you got a billion $$ probe on another planet, it surviveing is more important then a fast boot time heh. and you can always login and tell it to continue booting
"We discovered a system log in which the problem was documented,"
Those guys are running a very expensive experiment, are logging it and they have no idea what and where they are logging??
Slashdot: stuff for news, nerds that matter, matter for news, stuff that nerd
I did read the article, and my comments are completely accurate. Unfortunately you must not have made it to the 3rd paragraph, and neither did the mods that modded you up and me down.
The problem was discovered after launch. The first few fixes made the problem worse by stressing the filesystem even further.
It doesn't matter that they were trying to fix the problem. THAT WAS NOT MY POINT. The problem should have been identified and fixed before the craft was launched.
Yes, they may have taken "around" 100000 pictures. Does that mean they sequentially stored every picture in an actual rover file system? I get the impression they were only testing the cameras or the capture software, not the holistic system.
Did they first simulate filling the filesystem with files generated during the actual trip to mars? Apparently not, because the system would have failed if they had actually put the rover software through a launch to end of mission simulation here on earth when the software was developed.
Dan East
Better known as 318230.
Seriously, from a developer viewpoint, that is all wrong.
I have worked on projects in which there was simply too much logging going on that you couldn't tell head from toe anymore. When a problem arrived, scanning the logfiles proved very cumbersome indeed. Every developer had his own stuff logged, which sometimes proved interesting, sometimes proved utter crap (noone wants to know variable XYZ is increased by 1 for 24943 times).
You should develop a well-thought logging strategy that increases the logging verbosity on a problem-basis, not simply log everything that happens and hoping you get some useful information.
Slashdot: stuff for news, nerds that matter, matter for news, stuff that nerd
Not lost forever, but lost until we travel to Mars and retool it as an extraterrestrial barbeque grill.
"I'll say it again for the logic-impaired." -- Larry Wall.
...and I'm not saying that just because we agree. Yours are good additional insights (hence your "insightful" mods up! :-)
:-( "
:-)
I agree with the reply-post below too, saying that if they'd made their system a bit more fault-tolerant, then the problem might have been more easily recovered from. Sixty reboots in a row in a day seems a little excessive! Don't they have counters to detect that very thing? Don't they have a failsafe/debug OS burned into ROM (not flash) to load automatically in just such an event? Such are the risks when you're reloading a whole new OS remotely!
However, maybe they do have such things, or equivalent. I don't think their method of recovery was "accidental" (or a hack) either, although I'm making assumptions and I haven't seen their spec. The key is that they recovered from the error... and I now assume that they have recovered completely.
What I found interesting was NASA's initial assessment that the flash ROM was failing -- a hardware failure. The media jumped all over that and reported it, so the rest of us were thinking, "Great, the rover is crippled and will never be the same.
Now, turns out it was just a software error. Where's the mainstream media now? ("EE Times" is hardly mainstream!) Can the rover's recovery now be considered a "complete recovery"?
If this story goes mainstream, will it make NASA look bad for screwing up... or look good for making a full recovery? I'm not sure. (Of course, smart people make mistakes too, lots of them, but the key to being smart is covering your ass beforehand!
> What on earth (or on Mars) could we possibly take away from this experience?
Rule 3: Never ignore the return value from open.
well, this presupposes that what caused the problem in the first place also didn't mess up the hardware watchdog as well.
Nothing's perfect. It also presupposes that the sun didn't explode and vaporize the Earth and that God didn't get ticked off and squish it with his thumb, So What?
A watchdog is a VERY simple device. A simple countdown timer, a control register with associated address decode, etc. It's quite unlikely to fail. When the timer hits zero, it strobes reset. Any access to the port address resets the countdown timer.
Some dual processor boards are even set up to alternate which is the boot processor, so they can come up with a single failed CPU.
There is always some sort of problem that precludes recovery. No amount of software or clever design can help you if the device is destroyed. However, that doesn't mean don't even try.
It's not that hard to pull off off this sort of seemingly amazing remote recovery with pure off-the-shelf tech if you plan for it in advance and are willing to pay a modest premium.
You need remote serial console access -- ideally including firmware/bios serial console access -- and remote power cycling, controlled by a small embedded system, either in separate units (APC masterswitch, terminal servers) or as part of the system unit (common on Sun gear as "LOM"/"ALOM"/etc.; some of this is also creeping into x86 mobos). All this lets you regain control of the system remotely.
Then it becomes a matter of hardening the system to let you recover from various other insults. Never let go with both hands: Mirrored disks (protecting against hardware failure) and multiple bootable partitions (protecting against software or human error) can both be used; netbooting is also a nice capability to have when you've got a bunch of servers in the same place.
Disclaimer: I bet you can do much of the above with other people's gear, but I work for Sun and I know it works for me...
The enroute time for Cassini to get to Saturn was 7 years; rather than push back an already long mission they launched with feature-incomplete code. They knew they had 7 years to get the software fully functional and debugger; they've updated it remotely from millions of miles away a number of times now.
I'm sure the rovers did the same thing... Develop the launch/cruise software before you launch (and of course try to get as much of the entry/landing code done as you can!), and then uplink the final code before it's needed. Therefore it doesn't surprise me one bit that the JPL engineer knew there were shortcomings in the launch software.
Hell, I develop BIOS for servers and we do it all the time. The BIOS image we give the hardware engineers for initial bringup is usually *way* short of features that will be there when it actually gets used by the customers!
--Rob