EC2, S3, and JungleDisk

Though EC2 and S3 have both been around for almost 2 years now, I've only just recently got around to checking them out. First impressions, EC2 is great! Few minutes of work and you're got yourself a virtual machine running RH Linux. The vm looks and feels like any other remote machine. That said it is a virtual machine so if you want to presurve your work you'll need to do something. In my case it was decided to store all the important data on S3.

With S3 you don't get direct access to a disk. Instead you need to invoke a AWS (Amazon Web Service) to store, remove or retrieve data. In my case I was to use JungleDisk, a product that installs its self to look like a local device. Unfortunately this product sounds much better than it currently is.

The end game was to setup the virtual machine so that several people could login to it. JungleDisk by default only allows one user to access the mounted device. Since you need to be root to mount a device in Unix... And of course this wasn't so desirable. My first idea was to, mount the JungleDrive using the "-o allow_other" option. After that I would create the users home directory on the JungleDisk. I started jungledisk with the requiste options only to discover that FUSE didn't recognize the allow_other option as it was presented to it by jungledisk. Strangely enough, jungledisk went and created a -o directory and then complained that it couldn't mount allow_other! Worse than that, it went and modified the .ini file so that the -o mount point was now a permanent fixture.

Needless to say it took me quite some to sort out that the .ini file was being modfied and much longer to figure out how to fix it. While splunking through the forums I came to realize that I was going to have to abandon my initial plans. Instead what will happen is that root will gather things up that need to be preserved and it will then copy them to S3 using the single user mode of JungleDisk. Single user mode seems to be the only use case that is not full of show stopping bugs.

I fail to see why JungleDisk has choosen the model that they have. The software looks half like a webserver and half like a device driver. In looking half like both is seems to do each half as well as is needed. An alternative would have been to build something like NFS. NFS looks exactly like a local disk. It mounts like a locak disk and responds to admin commands like a local disk. On the other hand, JungleDisk makes the magical file system entries that look like a directory yet don't react to admin commands as one would expect them to. Yet this is but the least of the problems for this product which is a shame.

The shame is JungleDisk interfers with what would otherwise be a wonderful user experence provided by EC2 and S3.