MAGNET WEEKLY CTF #5
The question refers to the original filename of a block ID.
The relationship between filenames and block ID's in Linux was not readily apparent to me, even after a Google search which led me to a SANS presentation by Kevvie Fowler which involved using hdfs terminal commands (only viable after loading Hadoop tools on Linux) and perfoming a "checkpoint" to ensure accurate results...maybe there's an easier way.
My next stop after loading the file HDFS-MASTER.E01 into Autopsy was to check
/home/hadoop/.bash_history to see if a file had been copied and enter
the original filename as the flag.
mapred-site.xml.template was not the answer, however.
Fortunately, I had multiple tries at the answer this week.
A keyword search in Autopsy for the block ID:
Turned up this related file:
Which was not the answer, either.
It could be time to spend 20 points on the hint.
It was a link to the same Kevvie SANS presentation that I had already viewed.
Maybe I should watch it again, more closely.
Along with the technical Hadoop info,
he mentioned an Edit Log:
It looks like the required information is there,
but where is the Edit Log? Could I find it without mounting the image in Hadoop?
Stack overflow showed the possible location in the file system:
but there was no hadoop folder in the HDFS image /var/lib.
So, where are the logs?
An Autopsy keyword search for log led me to an
enormous list which I scrolled down looking for hadoop-0.20.
I stumbled upon another log named
hadoop-hadoop-namenode-master.log
Master name log looked promising.
When I opened it in Notepad...
...and searched for block ID 1073741825, I found this:
which was the flag, but it was too late to try it.
Lesson learned: look to the LOGS!
I'm looking forward to the other writeups so that I can gain more familiarity with Hadoop and maybe even try installing an instance so that I may try out the hdfs command line tools.
Comments
Post a Comment