12.16.09
Posted in Uncategorized at 2:58 pm by Nate Smith
Many people are suspicious of the reported trick of freezing a hard drive in a last ditch effort to recover it. Freezing a hard drive to recover data really can work. I don’t know why it works, but I have used it successfully on a disk that would not boot and I can tell you that heat is absolutely an enemy of hard drives. In a nutshell this is how it is done:
You remove the dead or clicking hard drive from the machine. You obtain a second hard drive of equal size or larger. Once frozen you restart the hard drive and quickly copy off all your important data or use dd, ghost, or some other tool to copy all the data over to the second hard drive. You might also hear about using tools like SpinRite to try to repair the damaged drive to get it to a state where it will work under normal or non-frozen circumstances. I prefer to consider the drive damaged goods and get all the information off it and then retire the drive.
There are some important things to know in order to freeze a drive for recovery:
- You need to keep moisture away from the hard drive. Seal the hard drive in an antistatic bag. Use a dessicant in the bag if you have one. You want to keep moisture away from the drive as much as possible. Seal the antistatic bag inside one or two other plastic bags reducing the amount of trapped air as much as possible.
- The drive has to freeze a long time. This is not an instant process. The drive needs to be really cold. Leaving it in the freezer for a 24 hours or more to get the drive chilled completely is a good practice.
- You have a limited amount of time to recover the disk becuase it will warm up fast. Really fast.
- If you need more time, you can buy dry ice at a grocery store and put the drive and dry ice inside a cooler. This is what I do when I recover with some of the linux-based dd tools because they take so long to run - on the order of days for a really big drive.
On new hard drives larger than 350 GB I try to make sure the drives have additional cooling installed. Extra fans or heat sinks to cool them down. I also like to do an initial format of the hard drive and let it warm up, copying some data to and from it to warm it up to “burn in the drive”, before finally formatting it. I do this under the theory that once warmed up the platter sizes are probably different from their cold size and that there is probably a point where the warmed up drive will not get any hotter unless it is under stress. Testing the hard drive before putting it into use also gives it a chance to fail or throw SMART errors before you put anything important on it.
You can freeze a hard drive to make a final try at recovering it but you need to work fast to keep the drive cold and dry. Sometimes a dead drive is truly dead and if it contains critical data or programs it must be sent to professionals to try advanced recovery steps.
Permalink
09.18.09
Posted in Semantics, Uncategorized, rant at 8:43 am by Nate Smith
There are a lot of terms that leave people scratching their heads but ‘cloud computing’ is one of the worst in a while. Everyone you talk to in and out of I.T. is thinking something different when the term comes up. The term ‘cloud computing’ is nebulous. (That is a good pun and I am not ashamed of it)
So let’s rectify this a bit. When we think of things like Google docs, Google mail and salesforce.com, we should call this the web application cloud. There are some one-trick poinies like Carbonite that work in the cloud too and may or may not use the web as their transport. These could be called Cloud products.
Then there are services we should think of as cloud computing services. Services like Amazon.com’s S3 storage are not a complete a product in themselves but are building blocks. These services are cloud computing services or web cloud computing services if they have only a web API (application programming interface).
Now it gets a bit tricky with vendors like VMware talking about building your own cloud. You can use products like VMware and others and build an on-premise cloud. This is multiple VMware servers or the Vsphere product letting you move inbstances of servers between physical machines. You might even be able to move them over a WAN or VPN to a remote hosting site or anyther company location. This might be the on-off-premise cloud.
It is possible to outsurce all your computers to someone else and run everything in their cloud of physical or virtual machines. This might be called outsourced infrastructure cloud or oursourced virtual infrastructure cloud.
Some people seem to imply that moving their operations or computer room contents to another company’s datacenter constitutes cloud computing. So let’s be inclusive and call this co-hosted cloud computing. When you do this it is usually done to get away from the issues of building and maintaining a computer room with special fire, cooling, and electrical services.
I’m not a big fan of many of the cloud computing types above. Even if you master the subleties of the financial arrangement, have goot toos to migrate and extract data, and even if you have a great exit strategy, the speed of light and fiber-seeking backhoes are still some of your worst enemies. When you boil it down, nearly any form of cloud computing is really outsourcing and the key is having a contract or agreement that is air-tight, and has rewards and penalties.
Permalink
01.19.09
Posted in AoE, Linux/*BSD/Unix, System Administration, Uncategorized at 9:51 am by Nate Smith
This is really written as a reminder to me since I seem to have to go in search of this information whenever I need to format large disks.
Using parted for GPT.
The parted tool is very utilitarian but it gets the job done.
GPT is necessary for disks with space probably greater than 4 Terrabytes. (I can’t remember if it is 2 or 4 or 4.5) but it would probably be a good practice to start using it on anything over 2. GPT replaces the old style partition tables. The old FAT and BIOS restrictions are some of the last vestiges of 16 bit computing cruft we carry along with us into the 64-bit world.
The version of parted I was using didn’t have support for XFS. This example is borrowed from the Coraid web site:
shell # parted /dev/etherd/e0.0
(parted) mklabel gpt
(parted) print Disk geometry for /dev/etherd/e0.0: 0kB - 3001GB Disk label type: gpt
(parted) mkpart primary 0 1000G
(parted) mkpart primary 1000G 2000G
(parted) mkpart primary 2000G 3001G
(parted) print
Disk geometry for /dev/etherd/e0.0: 0kB - 3001GB
Disk label type: gpt
Number Start End Size File system Name Flags
1 17kB 1000GB 1000GB
2 1000GB 2000GB 1000GB
3 2000GB 3001GB 1001GB
(parted) quit shell
# cd /dev/etherd/ shell
# ls e0.0* e0.0 e0.0p1 e0.0p2 e0.0p3 shell
# mkfs.ext3 /dev/etherd/e0.0p1 shell
# mkfs.xfs /dev/etherd/e0.0p2 shell
# mkfs.reiserfs /dev/etherd/e0.0p3
in this case the disk in question is an AOE disk, but if you had a large SCSI array you would just substitute that instead of /dev/etherd/e0.0.
Parted can make the filesystem (mkfs) for you, type “help” at the (parted) prompt. EXT3 was not implemented in the version I was using.
`tune2fs -l /dev/sda1 | grep UUID” is how I find the UUID of the drive. If there are other ways to find the UUID of a drive I would like to learn them.
once you know the UUID for a drive you can mount it using just the UUID using the -U flag in the mount command:
mount -v -U 2ab524bc-3640-4dfd-bf4c-0373172e0159 -t ext3 /mnt/ether/
in the command above
-v yeilds verbose information about the mount command
-U tells the mount command to use the UUID of the device and is followed by the one we want to use. (those UUIDs are long aren’t they?)
-t tells the type of filesystem we are mounting. In this case, ext3. This is not always necessary
finally /mnt/ether/ is the directory I have set up for testing this etherdisk drive.
You can add the entry to fstab to cause a mount. Just remember if you are using aoe to make sure the aoe module is loaded first! Other large disk types won’t have that problem. This is what the entry should look like:
GUID=a241eee7-13e7-425d-854e-fa5ec1ee4c62 /mnt/largefs auto defaults 0 0
Now we have great big open file systems to save all that crufty old data on.
Permalink
11.07.08
Posted in Uncategorized at 4:27 pm by Nate Smith
One thing that makes me very happy is the sight of parts of wind turbines on semi trucks on the highway. Lately a lot of wind farms are being built in Iowa and I think it is one of the greatest things to happen in my lifetime. I don’t know much about the lifespan or the maintenance on one of these behemoths, but it seems like such an obvious idea when you compare it to someting like a coal or nuclear power facility. Wind power is distributed, not inherently dangerous, and the raw material (wind) supplies itself; there also isn’t a giant ecosystem set up around it like some other production methods. I’m not foolish enough to think that wind can supply all our needs, but it sure is great to finally see it being explored in a meaningful way.
Permalink
04.10.08
Posted in Uncategorized at 10:42 am by Nate Smith
We have a little application that collects the contents of a database and then sends them to a few mail recipients for review. It was working fine in CF 8 prior to moving to Exchange 2007. We have other applications that relay mail through Exchange 2007 fine but for some reason this one was not. It turned out that the cfmail from field only had the name and not the domain of the sender. Once we added the sender domain mail would relay and work again.
This is probably a good thing - Exchange having tighter adherence to a standard, though I’m not sure which one. But it could definitely be a “gotcha” if your mail server was working fine prior to a switch to Exchange 2007, or possibly even an earlier version of Exchange such as 2003 or 2000. We happened to be moving away from Exchange 5.5.
Permalink
01.05.08
Posted in Uncategorized, miscellany, ramblings at 10:10 pm by Nate Smith
Some interesting things have gone on with regard to the organization of information during the evolution of the World Wide Web. Initially, the ability to link relevant items was key to the Web. While this is still important and the basis of the Web, search engines have almost replaced linking as a means to locate information. The search engines themselves require the links so links will always remain the key organizational feature of the Web.
The development of the Wiki realized the promise of sir Tim Berners-lee’s original vision of the web — at least until the wiki-spammers arrived to crush it. A wiki has micro links on an organized site. Relevant information tightly linked and easily updated. One might argue Wikipedia is almost the antithesis of what a wiki should be. This is because the scope of Wikipedia is really too broad for a classic wiki. Wikipedia requires special disambiguation entries in order to distinguish between terms that overlap subjects.
Then came the blog. Blogs are mostly comprised of chronologically organized, granular, topic-focused entries. Of course blog entries usually have their fair share of links to other pages. Initially few blogs had original material and were in fact a chronological pointer to interesting things happening on other web sites.
The social network is another interesting facet of information organization. Sites like facebook and linked-in organized on the basis of social connections.
Now we have immediate information organized by the likes of twitter. It will be interesting to see if anything useful comes of that.
My own idea is to combine the highly linked and organized nature of a wiki with the chronology of blogs and twitter. Wikis have a problem where old information goes stagnant and does not get removed or updated, but there is no way to easily tell what is new information and what is old unless authors go to lengths to make it apparent. I would like to see a wiki - I call it a horizon wiki - where each entry has a date or timestamp associated with it. Old entries would be visually different from new entries and corrections. For instance new entries could be dark while older entries are progressively lighter, or vice versa. It would be visually apparent what information was recent and what was old so the reader would have a cue to make sure it is still relevant.
There are a lot of neat experiments with information organization going on right now too. One of my favorites is the news application on the Nintendo Wii. You can place it in a mode where it shows a globe. Places that have news stories show as a small thumbnail picture. The globe is simulated 3D so you can rotate it and see that some locations have stacks of stories with a number to indicate how many while some only have one or two items in the stack stories. This is a neat combination of geographical, and chronological organization - older stories are on the bottom of the stack. Innovative.
Permalink
09.30.07
Posted in Uncategorized at 10:45 am by Nate Smith
The latest version of KnoppMyth (http://mysettopbox.tv) has support for Schedules Direct new service. Knoppmyth is easily the best way to get MythTV up and Running on just about any hardware. I had it installed and running for about two and a half years on an older version. It’s software I really *used* everyday. It works best with nvidiavideo drivers but will work with anything. I have it setup with a front end and a back end as described on the knoppmythwiki. Some of the greatest additions to MythTV are already installed and working on KnoppMyth.
Permalink
03.07.07
Posted in Uncategorized, miscellany, ramblings at 1:05 pm by Nate Smith
As people get older I think they begin to question what reality they are a part of and wonder if it is the same one they share with others. I believe this on an uncouscious level. You can see it in everyday conversation, people comment on wether it is cold, or hot, or the particular sharde of green the grass is. It’s like they are trying to nail down a common experience that they are sharing.
Example:
“That was very loud, did you think that was loud? I thought that was very loud.” Translated: I just had an experience, something very loud - did you have the same experience I did?
There are probably several reasons this dissonance appears as people age. Their senses begin to change, their ethics change, they might experience chemical induced experiences - all these cause people’s sense of existance to vary.
More to come.
Permalink
03.10.06
Posted in Uncategorized, paranoia at 12:39 pm by Nate Smith
Chicken McNuggets are strange. Little toasty shapes of greyish fluff with a taste similar to chicken. I have a theory about their source but, like most good *theories* it cannot be proven.
Chicken McNuggets are really some kind of genetically engineered chicken-flavored mushroom. They grow to exactly the right size and can be coated with batter and fried up as a McNugget.
I’m probably being paranoid, but if you get to see one again pull it apart and look at the “chicken”. I think you will agree there is something suspect about it.
Permalink