While the web interface on the device isn’t too bad, I wanted a little more control over it, as well as the ability to monitor the device’s health with some custom munin graphs.

Here’s how to enable SSH without wiping the drive and installing custom firmware.

1. Set up a new user account on the device from the web interface, and create a password.
2. Shut down the device, remove both drives, and connect them to a local computer with SATA cables.
3. Boot your linux computer with the drives attached, it’s probably safest to boot from some sort of Live CD. I used Fedora.
4. When you get to a shell, execute the following commands to mount the drives.
   $ mdadm -A /dev/md0 /dev/sda1 /dev/sdb1
mdadm: /dev/md0 has been started with 2 drives.
$ mkdir /mnt/md0
$ mount /dev/md0 /mnt/md0

5. Add the following line to the file /mnt/md0/etc/inetd.conf:
   ssh stream tcp nowait root /usr/sbin/dropbear dropbear -i
   You can do this in one shell command by executing this:

   echo "ssh stream tcp nowait root /usr/sbin/dropbear dropbear -i" >> /mnt/md0/etc/inetd.conf

6. Edit /mnt/md0/etc/shadow with a text editor, and replace the hashed password for “root” with the hashed password of the user you created. The hashed password is the characters between the second two colons, and begins with $1$$.
7. Unmount the drives and shut down the array:
$ cd
$ umount /mnt/md0
$ mdadm -S /dev/md0
mdadm: stopped /dev/md0
8. Put the drives back in the NAS device and turn it back on!
9. After it boots, you’ll be able to ssh in as root with the password you set from the web interface.

Here is a snippet of my virtual hosts definition for geoloqi.com:

<VirtualHost *:80>
	ServerName geoloqi.com
	ServerAlias *.geoloqi.com
	ServerAlias geoloqui.com
	ServerAlias *.geoloqui.com

	RewriteEngine on
	RewriteCond %{HTTP_HOST} !^geoloqi.com
	RewriteRule ^/(.*)$ http://geoloqi.com/$1?utm_source=DomainRedirect&utm_medium=%{HTTP_HOST} [R=permanent,L]
...

This will make domain typos visible in Google Analytics under the “All Traffic Sources” section. You can clearly see the referrals from the typo’d domain name at the bottom of this list.

You could also use the filter box to search for “DomainRedirect” and see only the typos.

I’d be curious to know if anybody else has other suggestions for tracking this type of thing.

Dropbox - Online file backup and sharing

I’ve recently been having problems with Dropbox ever since joining a shared folder which overflowed my quota. Completely unrelated files mysteriously disappeared. Luckily I had a snapshot from before I had joined the folder and was able to restore the files.

This prompted me to set up a proper backup system which makes snapshots daily, weekly and monthly.

This guide assumes you have a Linux server somewhere with no GUI.

First, follow the Dropbox wiki instructions for installing Dropbox on Linux. The process is relatively painless, and seems to just work. At the end, you’ll end up with a “Dropbox” folder in your home folder. This folder will always have the latest version of your files.

The next step is to take snapshots of that folder at regular intervals. I chose to keep 7 days of daily snapshots, 4 snapshots taken weekly, and 12 taken monthly. In theory, this should give me access to any files I’ve accidentally deleted within the past year.

This script will run rsync to clone the Dropbox folder into snapshot folders which will be overwritten over time, so you will only end up with 7+4+12 copies of your Dropbox folder. Save this code to a file called backup.sh or whatever you want, and make it executable (chmod 755 backup.sh).

backup.sh:

#!/bin/bash
 
# Makes daily snapshots in folders like "daily-4-Thu", "daily-5-Fri"
if [[ "$1" == "daily" ]]
then
        path=daily-`date +%u-%a`
fi
 
# Makes weekly snapshots in folders like "weekly-1" where 1 is the day of the month
if [[ "$1" == "weekly" ]]
then
        path=weekly-`date +%d`
fi
 
# Makes  monthly snapshots in folders like "monthly-04-Apr"
if [[ "$1" == "monthly" ]]
then
        path=monthly-`date +%m-%b`
fi
 
# Run with "go" as the second CLI parameter to actually run the rsync command, otherwise prints the command that would have been run (useful for testing)
if [[ "$2" == "go" ]]
then
        rsync -avz --delete /home/aaron/Dropbox /home/aaron/aaron/Dropbox-Backup/$path
else
        echo rsync -avz --delete /home/aaron/Dropbox /home/aaron/Dropbox-Backup/$path
fi

Add these lines to your crontab to run the backup script at the appropriate intervals. This will run the daily script every night at 1am, the weekly script on the 4th, 12th, 20th and 28th days of the month at 1:30am, and the monthly script on the 1st of the month at 2am.

crontab:

0 1 * * * /home/aaron/backup.sh daily go
30 1 4,12,20,28 * * /home/aaron/backup.sh weekly go
0 2 1 * * /home/aaron/backup.sh monthly go

Hopefully this saves you some headaches down the road when Dropbox decides to play tricks on you.

Using Linux’s LVM, I can create a logical partition from the two smaller drives as big as the 1tb drive.

$ pvdisplay

Lists all physical volumes managed by LVM. First we have to create the physical volumes for LVM. I prefer to create the volumes out of partitions, although you can do it from raw drives too.

First let’s use fdisk to create “Linux LVM” partitions on the two drives.

$ fdisk /dev/sdb
Press "n" to create a new partition
Press "t" to set the partition type, and enter "8e" for Linux LVM.

When creating the partition, accepting the defaults will make it use the whole drive. I want to use the entire 750 drive and only part of the 320 drive, so that in total it has the same number of blocks as the 1tb drives. So I first created the “Linux RAID” partitions on the 1tb drives so I could see how many cylinders it listed, which ended up being 121601. So I created a partition the full size of the 750 drive (91201 cylinders), then created a 121601-91201 cylinder partition on the 320 drive.

$ fdisk -l /dev/sdb
Disk /dev/sdb: 320.0 GB, 320072933376 bytes
255 heads, 63 sectors/track, 38913 cylinders
   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1               1       30400   244187968+  8e  Linux LVM
/dev/sdb2           30401       38913    68380672+  83  Linux
 
$ fdisk -l /dev/sdj
Disk /dev/sdj: 750.1 GB, 750156374016 bytes
255 heads, 63 sectors/track, 91201 cylinders
   Device Boot      Start         End      Blocks   Id  System
/dev/sdj1               1       91201   732572001   8e  Linux LVM

Now that I have the two partitions ready, I moved on to LVM setup.

$ pvcreate /dev/sdb1
$ pvcreate /dev/sdj1

This sets up the two partitions as physical volumes for LVM.

Next is creating a logical volume group:

$ vgcreate vg_tb /dev/sdb1 /dev/sdj1

This creates a volume group called “vg_tb” using the two physical volumes sdb1 and sdj1.

Let’s take a look at what we have so far:

$ pvdisplay
  --- Physical volume ---
  PV Name               /dev/sdb1
  VG Name               vg_tb
  PV Size               232.88 GB / not usable 832.50 KB
  Allocatable           yes
  PE Size (KByte)       4096
  Total PE              59616
  Free PE               59616
  Allocated PE          0
  PV UUID               70hPKX-n11U-RcB6-0Kyt-1SOP-ni7E-2Y9hcE

  --- Physical volume ---
  PV Name               /dev/sdj1
  VG Name               vg_tb
  PV Size               698.64 GB / not usable 2.34 MB
  Allocatable           yes
  PE Size (KByte)       4096
  Total PE              178850
  Free PE               178850
  Allocated PE          0
  PV UUID               PzFb9b-lapG-KdT3-78nh-Gq75-F0Lo-I3xCrl

  --- Physical volume ---
  PV Name               /dev/sda2
  VG Name               VolGroup00
  PV Size               111.60 GB / not usable 2.86 MB
  Allocatable           yes
  PE Size (KByte)       32768
  Total PE              3571
  Free PE               1
  Allocated PE          3570
  PV UUID               1wM65Z-3QGd-vDiq-mq1R-YhEE-Ackp-hh3g13

$ vgdisplay
  --- Volume group ---
  VG Name               vg_tb
  System ID
  Format                lvm2
  Metadata Areas        2
  Metadata Sequence No  1
  VG Access             read/write
  VG Status             resizable
  MAX LV                0
  Cur LV                0
  Open LV               0
  Max PV                0
  Cur PV                2
  Act PV                2
  VG Size               931.51 GB
  PE Size               4.00 MB
  Total PE              238466
  Alloc PE / Size       0 / 0
  Free  PE / Size       238466 / 931.51 GB
  VG UUID               olg9GP-x1sC-sFAD-TgWY-KIIx-YWNt-kL763n

  --- Volume group ---
  VG Name               VolGroup00
  System ID
  Format                lvm2
  Metadata Areas        1
  Metadata Sequence No  3
  VG Access             read/write
  VG Status             resizable
  MAX LV                0
  Cur LV                2
  Open LV               2
  Max PV                0
  Cur PV                1
  Act PV                1
  VG Size               111.59 GB
  PE Size               32.00 MB
  Total PE              3571
  Alloc PE / Size       3570 / 111.56 GB
  Free  PE / Size       1 / 32.00 MB
  VG UUID               IZ25LV-oMOG-DwKK-QuFN-bqqp-ClUe-d71k5l

So far so good. The next step is to create a logical volume in the new volume group:

$ lvcreate vg_tb -n onetb -l 100%VG
  Logical volume "onetb" created

This creates a new logical volume called “onetb” in the “vg_tb” group using 100% of the group’s available space. Now let’s take a look at the list of logical volumes:

$ lvdisplay
  --- Logical volume ---
  LV Name                /dev/vg_tb/onetb
  VG Name                vg_tb
  LV UUID                sQKaq9-D6Mv-p8it-vWGW-O7DX-FGmC-cl5FSh
  LV Write Access        read/write
  LV Status              available
  # open                 0
  LV Size                931.51 GB
  Current LE             238466
  Segments               2
  Allocation             inherit
  Read ahead sectors     auto
  - currently set to     256
  Block device           253:2

  --- Logical volume ---
  LV Name                /dev/VolGroup00/LogVol00
  VG Name                VolGroup00
  LV UUID                G5hlbb-tA3S-qhTS-03us-f9dl-1Vxy-9vDSU5
  LV Write Access        read/write
  LV Status              available
  # open                 1
  LV Size                109.62 GB
  Current LE             3508
  Segments               1
  Allocation             inherit
  Read ahead sectors     auto
  - currently set to     256
  Block device           253:0

  --- Logical volume ---
  LV Name                /dev/VolGroup00/LogVol01
  VG Name                VolGroup00
  LV UUID                7NZrY9-1wSJ-4fRp-VPnM-V07u-9rj8-vmtxTx
  LV Write Access        read/write
  LV Status              available
  # open                 1
  LV Size                1.94 GB
  Current LE             62
  Segments               1
  Allocation             inherit
  Read ahead sectors     auto
  - currently set to     256
  Block device           253:1

You can ignore all of the VolGroup00 things, those are the auto-created volumes from when I installed Fedora.

At this point, I have a new device at /dev/vg_tb/onetb which is the same size as my 1tb drives, and I can use it exactly as I would use the 1tb partition at /dev/sdf1.

Now it’s time to create the RAID 5 array from these four volumes.

$ mdadm -v --create /dev/md1 --chunk=128 --level=5 --raid-devices=4 /dev/sdf1 /dev/sdh1 /dev/sdi1 /dev/vg_tb/onetb
mdadm: layout defaults to left-symmetric
mdadm: layout defaults to left-symmetric
mdadm: layout defaults to left-symmetric
mdadm: layout defaults to left-symmetric
mdadm: layout defaults to left-symmetric
mdadm: size set to 976756608K
mdadm: array /dev/md1 started.

The array will begin syncing, and you can watch it by running:

$ watch -n 1 "cat /proc/mdstat"
Personalities : [raid6] [raid5] [raid4]
md1 : active raid5 dm-2[4] sdi1[2] sdh1[1] sdf1[0]
      2930269824 blocks level 5, 128k chunk, algorithm 2 [4/3] [UUU_]
      [=>...................]  recovery =  5.5% (54017536/976756608) finish=593.1min speed=25925K/sec
 
md0 : active raid5 sdg1[0] sdc1[3] sde1[2] sdd1[1]
      2197715712 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
 
unused devices: <none>

While this is syncing, we can create the ext3 filesystem.

$ mke2fs -j -b 4096 -m 0 -E stride=32,stripe-width=96 /dev/md1

This creates an ext2 filesystem with journaling (ext3), the block size is 4kb, and 0% of the blocks are reserved for the superuser. The stride is calculated as the raid block size / ext2 block size (128k / 4k = 32). The stripe width is calculated as the stride value times the number of data disks in the array. In a 4-disk RAID 5 array, there are three data disks, one being for the parity data.

This will take some time, and will significantly slow down the sync process. Mine dropped from 25mb/s to around 1mb/s. I figure I’ll let it create the filesystem so I can start copying data to it right away, and let it finish its sync on its own time.

While you’re at it, you should set up munin to monitor the SMART data from all the drives as well as the status of the array using my munin-raid-monitor plugin.

Additional Reading:

The easiest solution I could come up with was to tell my mail server to listen on another port such as 587, and have my firewall route requests for port 25 to port 587. Here is the iptables rule to do that!

iptables -t nat -A PREROUTING -p tcp -i eth0 -d xx.xx.xx.xx --dport 25 -j DNAT --to-destination :587

Where xx.xx.xx.xx is the IP address of my mail server. Now all the computers inside think they are communicating with my mail server on port 25, but the firewall secretly passes the request on to port 587 instead.

Because sshfs on OSX’s version of sshfs is wonky.

My solution is to use an intermediate linux server which does the sshfs mount, then serves that to os x over a samba share. Another benefit to this is that you only need one samba share to mount all your sshfs connections, since the linux server will be taking care of those. Also if you had a windows computer you wanted to use this with, it would also be able to mount the samba share from the linux server.

Let’s get started.

I ran into a couple tricky config issues while setting this up. This post is here mostly for me to remember them for next time I need to do them. If anybody else happens to stumble across this and finds it useful, then that’s a bonus.

On the linux server, you’ll need to install fuse-sshfs as well as samba. In Fedora, this can be done like so:

$ yum install samba
$ yum install fuse-sshfs

You need to add the linux user to the fuse group, and create a samba user account.

$ usermod -a -G fuse aaron
$ smbpasswd -a aaron

Here’s something that isn’t normally covered in fuse tutorials. In order to allow samba to access the fuse mount, you need to create a file, /etc/fuse.conf with the following contents:

user_allow_other

If you don’t do that, the mounted folder just disappears from the samba share.

You need to make some changes to /etc/samba/smb.conf file in order for symlinks to be shared to os x. I’m not sure if this is required for fuse to work, but it’s nice to have anyway:

unix extensions = no  (add this outside of a share definition)
follow symlinks = yes  (add this inside a share definition)

Of course, you need to configure the firewall to allow access to samba, (tcp and udp ports 445, 137, 138, 139 should do). And, you’ll need to make sure samba starts when the machine boots.

$ chkconfig smb on

We’re almost there.

To actually mount the sshfs folder, you’ll run a command which looks something like this. Note the extra option at the end:

sshfs username@example.com:/home/username mount_target -o allow_other

Now you can mount your home folder on the linux server over samba, and you’ll see a folder mount_target, which is the sshfs mount.

Note: you’ll probably want to set up your ssh server with public key authentication so you don’t have to enter your password every time you connect. This is not the topic of this post, so I won’t bother mentioning how. There’s plenty of other tutorials on the Internet.

I hope this covers it, but feel free to comment if I’ve left anything out.

First let me just say that I have been struggling with this problem for quite some time now. The problem is to provide redundancy for a website so that the website continues to run even if there is a problem with one of the servers it’s running on.

In a typical simple server setup, there is a single machine running the web and mysql servers. The machine can be either a dedicated server, or as I have been using, a VPS. I have been running my websites off of VPSs for several years now, with minimal trouble. This works most of the time, but the having a site on must one machine means a Single Point Of Failure. If something is wrong with that server, the websites are non-functional until it is fixed. The trouble I have run in to falls under a few categories:

1. A problem with the physical host
2. A problem at the VPS level (operating system, Apache or MySQL errors, etc)
3. A problem at the network level

Problems with the physical host

Problems with the physical host do occur. With a VPS, these are completely out of your control, and are the responsibility of the hosting provider. Some examples of things I’ve encountered include a failed RAID array, a corrupted filesystem on the host, requiring a several-hour-long fsck, or an unplugged power cord. The worst issue I’ve had was when the provider said they had lost 2 drives in a RAID 5 array, and all the VPSs on that host were completely gone. Luckily I had a backup of the system and was up and running on a new VPS within a couple hours.

Problems at the application level

I haven’t actually run in to many problems at the VPS level compared to the other types of problems. However the latest issue I’m having does fall in to this category. Apache periodically starts crashing part way through serving a page with the error “[notice] child pid 21106 exit signal Segmentation fault (11)”. Visitors see a completely blank page some of the time.

Problems at the nework level

By far the most frequently occurring problem I encounter is network-related. These problems are usually out of both my and the hosting provider’s control. If there is a problem with the network, the downtime can vary greatly, anywhere from 5 minutes to 12 hours. It can be caused by a Denial of Service attack on a completely different server in the same datacenter, or it could just be a routing issue somewhere along the path from me to the server.

A typical redundant setup will cover both #1 and #2. Typical setups may include one or more load balancers in front of multiple application servers. If a machine goes down, the load balancers can stop sending requests to it. This works great if you’re trying to protect against servers failing. However if the load balancers are all on the same network, unless the network has multiple redundant paths, the whole system is still vulnerable to network issues.

My Solution

Since I most frequently encounter network issues, I can’t get away with a just a typical load-balancing solution. What I really need is a copy of the entire website in a geographically different location. Here is my solution:

One VPS in Dallas, TX (called triton), and another VPS in Newark, NJ (called proteus). (Yes, I name my servers after Greek mythology.) Triton holds the master copies of the websites’ php files, and proteus gets a copy of them via rsync. If I ever need to update the site, I edit the files on triton and then rsync them to proteus. Here is where the redundancy comes in. My DNS entries point the domain to both IP addresses. This means during normal operation, visitors will be more or less distributed between the two hosts evenly. If one server goes down, I can stop resolving DNS queries to it, and the worst that will happen is some dead pages for as long as the TTL on the domain.

This works as long as you’re just serving static content. Serving dynamic content, such as from a database, gets a little more complicated. MySQL’s NDB clusters are apparently only effective when run within a high speed network, with at least a 10 MBPS connection between them. Replication turns out to be more along the lines of what I’m looking for.

Replication to the rescue!

Replication is designed for a one-way sync between a master and slave. However, it is possible to configure two servers to be both a master and a slave. They will both notify each other of changes made to their databases. There is one trick you need to do in order to prevent primary key conflicts if rows are written to both databases while the link is down. It involves setting the auto_increment offset and increment, so that one server will only create even keys, and the other creates only odd keys.

/etc/my.cnf
    auto_increment_increment = 2
    auto_increment_offset = 1

Here’s some dry reading on replication from the MySQL manual. Here is a slightly clearer guide to replication which sums everything up pretty nicely. Overall, replication was pretty easy to set up. It seems to be pretty robust as well. I simulated network problems by adding firewall rules to block the servers from each other. I was able to continue to interact with each database, and the changes were all carried over when the link came back up.

Feel free to comment if you have any experience or insights into configuring web and database servers! I’m curious to hear what other people have done.

At some point, the array was “started” but had two faulty drives. I tried –remove and –add to remove and re-add the “faulty” drives. This had the effect of bringing the array back “online” with all the drives as spares. I removed the drives again, and tried the trick I used last time:

mdadm --assemble -f /dev/md0 /dev/sda2 /dev/etherd/e4.1 /dev/etherd/e4.2

However, this also didn’t work. It showed the array with /dev/sda2 and /dev/etherd/e4.2 as spares, and e4.1 was nowhere to be seen. At this point I was a little more than worried that I had done something to trash the array. That’s when a google search led me to this handy command:

mdadm -E /dev/sda2

This prints out the superblock information that is present on the hard drive. This told me that the e4.2 drive had not been damaged, since I was able to see information there. Also, the UUIDs on all three drives still matched. However, the bottom section of the report differed on all the drives.

A few google searches later, and I came across this:

mdadm --create --assume-clean --level=5 --raid-devices=3 /dev/md0 /dev/sda2 /dev/etherd/e4.2 /dev/etherd/e4.1

Using the –assume-clean flag tells mdadm not to write any data to the drives, or to start initializing the array. However, what I didn’t realize was that it would reset the UUIDs. That command brought the array back online, at least according to /proc/mdstat, but when I tried to mount it, it couldn’t figure out the filesystem.

That’s when I realized that the *order* in which you specify the drives to the –create command actually matters. I re-ran the command like this:

mdadm --create --assume-clean --level=5 --raid-devices=3 /dev/md0 /dev/sda2 /dev/etherd/e4.1 /dev/etherd/e4.2

The array came back online, and I was able to mount it!

So while RAID 5 protects against a single hard drive failing, it does not protect against me running stupid commands on the array. I’m going to have to start backing up my raid arrays onto other drives…

Useful Links

• http://ubuntuforums.org/archive/index.php/t-410136.html
• http://www.linuxdevcenter.com/pub/a/linux/2002/12/05/RAID.html?page=2
• Managing RAID on Linux (Amazon.com)
• http://episteme.arstechnica.com/eve/forums/a/tpc/f/96509133/m/697004072931

Question 2: File sum and multiplication problem

In this problem you are given a zip folder with hundreds of text files nested in folders an arbitrary number of levels deep. The question asks to find the sum of the nth line of all files with abc in their path, ending in .xyz. You do this twice, then calculate the product of the two sums. You can download the file I got from them at the link below.
googletreasurehunt.zip

My question is below:

Sum of line 3 for all files with path or name containing jkl and ending in .rtf
Sum of line 5 for all files with path or name containing stu and ending in .rtf
Multiply all the above sums together and enter the product below.

I thought this would be a good chance to practice my bash scripting, so I decided to give it a shot.

$ find . -type f -wholename "*jkl*\.rtf"

-type f restricts the search to files, not directories. -wholename searches for files matching that pattern. Once we run this, we get a list of all the files we’ll need to process. Now we need to create a running total of the values in the nth line in each file. This is where a bash ‘for’ loop and ‘awk’ come in handy.

$ for x in $(find . -type f -wholename "*jkl*\.rtf")

The ‘$’ tells bash to execute that command and return the result. The ‘for’ loop looks at the output line by line, and the variable x will contain each filename.

$ awk 'NR==3' filename.txt

This is an awk command that prints the 3rd line of filename.txt. If we can stick this into the ‘for’ loop, we can get a running total.

sum1=0
for x in $(find . -type f -wholename "*jkl*\.rtf")
do
        sum1=$(( `awk 'NR==3' $x` + $sum1 ))
done
echo $sum1

Now we just repeat for the other pattern, and we end up with our sum in two variables. Calculating the product is simple.

echo $(( $sum1 * $sum2 ))

Personally, I thought this puzzle was a little bit simplistic. I would have expected them to throw in some tricks, so that a poorly written search command would return some other files that would mess with your result. Now that I’ve said that, I hope I got it right! I find out in 10 hours if my answer is correct.

Feel free to comment if you have a different and/or better solution to this problem!

wait_timeout=604800
interactive_timeout=604800

That’ll do for now. The other alternative would be to check if the connection is alive before running a query, and connecting if it isn’t. I couldn’t get that one to work, though.