I recently flew on Continental 23 from Dublin to Newark.  This was the first flight where I experienced the new policy by Continental to only accept credit cards for drink purchases.  Funny enough, I was sitting in the first row and was the first passenger to use it – and it broke.  The rest of the flight, the crew had to fall back to cash only.  This was troubling to some as Continental had sent out a global e-mail and twitter announcement that they were only going to take credit cards.  I imagine some people probably skipped the stop at the cash machine!

Luckily I had tweeted asking and Continental had tweeted back that they still accept continental currency.  So, the exclusively accepting credit card thing, wasn’t so exclusive after all.  As a result, I had brought a couple of continental currency coupons and also had plenty of cash.  So, I was relatively unaffected by this occurrence.  Still, it lead me to ponder on my 8 hour flight from Dublin to Newark the use cases of only accepting credit cards.

I asked the flight attendant if they could run tabs.  He said they could, but they generally preferred to just close out after each purchase.  That better be one fast credit card machine.  I also sincerely hope they don’t print a receipt each time.  Thinking about this more, the entire concept just seems inefficient.  Sure, it appears to solve the problem of the flight attendants always having to find change w/ cash purchases.  However, it seems as though the credit card machines just create a new set of problems.

One wonders why you even need a credit card.  Shouldn’t you be able to just “sign for the purchases” like you would at a hotel? When you buy your ticket, you can essentially setup an account and the final bill is paid when your flight is over.  This would be incredibly convenient and would address a majority of passengers.  Those that booked – not on continental.com – could still present a credit card; however, it should only need to be swiped once “to cover incidentals.”  Even better yet, why would this account simply not be setup via the on-screen display?  To take it a step further, why can’t I just put in my drink order via this same display?

In short, having to process a credit card for every on-board purchase seems like a nightmare and I would bet will result in diminished service.

All your status belong to us!  For quite a while I have been intrigued by the p2theme theme from the folks at Automattic.  It also bothered me that all these status updates, were stored on twitter’s servers and at the whim of the twitterverse.   This past weekend, something happened that changed everything.  Wordpress decided to adopt the RSS Cloud protocol so that all word press blogs would be cloud enabled for a real time feed.

I talked to some of our developers and immediately we got to work on connecting a p2theme blog, status.jpmaxman.com, to my twitter feed.  It worked brilliantly.  Now, I can do all my status updates on my p2theme blog and they will instantly make it over to twitter.  From there I have a conduit to facebook & friend feed.  But, the origin feed, the master feed if you will, exists on my server and the content stays with me.  Should twitter get acquired and start inserting massive ads or get driven into the ground, I still have my status history.  In addition, as more and more services adopt the cloud feed, people can actually just subscribe to my conduit.  It can be a 100% peer to peer distributed twitter.  Dave Winer has been on a soap box about this for some time, and I’ve always agreed with him.  The Internet is meant to be distributed!

After thinking about this some more, I realized this goes way beyond status updates.  Any site – in particular news sites which we happen to run – could benefit greatly from a cloud feed conduit into other social media.  This conduit could update twitter feeds, facebook profile or fan pages, etc.  As stories are published to the news site they could instantly be syndicated to any number of places across the Internet.  As such, I decided to register: instafeed.net and put a public face to this.

We will be setting it up as an open source project and will be soliciting any help we can get in building this out.  Right now we are working on some top level architecture schemes as well as improving the core for an initial launch that will essentially just be RSS Cloud to twitter.

This is one of those projects that two weeks ago I would have never thought I’d be this deep into.  But, the pieces just kept coming together and it was too tempting to not see where it lead.  Contact me if you are interested in getting involved!

This is the second part to my previous post “Migrating Zimbra to Rackspace Cloud from Dedicated Rackspace Server” – I decided to put up separate posts as the first one was already pretty long.  After receiving the updated instructions I uninstalled the existing installation and set out to start over.    One cool thing about the cloud is when I decided to start fresh, through my control panel, I can just rebuild the machine to the fresh install.  I did this and set out to migrate again.  I ran into several problems along the way, but they were mostly just random things – some having to do w/ migration and some just with installing on Ubuntu.  In case anyone else runs into them I’ll outline the issues and the solution:

• LDAP time out errors
  Since I was migrating from one server to another, I wanted to get the second server setup before taking the other completely offline.  This meant that the DNS for the hostname was still technically pointed to the old server.  Once I manually put in an entry in the hosts file pointing the mail server domain name to the local public IP address, the timeouts went away and the install completed normally.
• libpcre3 missing
  During the install process it complained, though it did not error out, about this missing library.  Simply doing an apt-get to install it fixed this error.
• Helpful Hint – Logins / Password from the old server
  One thing in the instructions provided that you need to do is get your LDAP / MySQL passwords and your accounts used for spam / not spam.  To do this, on the old server:

  zmprov gacf | grep -i account

  And to get the ldap passwords:

  zmlocalconfig -s | grep -i password

  Much thanks to ZImbra support for that tip!  Made things much easier.

• SMTP Freezes after MAIL FROM
  The install and migration of the files went relatively smoothly.  When I fired things backup I did a manual telnet to port 25 to check that it was receiving mail.  The SMTP server would just freeze after I issued the MAIL FROM.  I decided to take a page from the other migration guide that said after you had finished migration to re-run the installation script.  After doing so this problem cleared.
• parts_decode_ext FAILED: Unix utility file
  Finally mail was being accepted, but I noticed it was all getting deferred.  After looking at the error message, “zimbra parts_decode_ext fail,” and doing a search – it turns out it was missing “file” a command line tool.  Again using apt-get to install file cleared this up.
• Cleaning up
  I then had to do some random clean up stuff, like adjusting the IPs that were allowed to relay to include the IP of the mail server.  Finally, some mail was getting deferred because it was trying to deliver it to the old mail server.  Once I finally switched the DNS (I had put the TTL down to 5 min) this cleared up.

So, I now have a successfully running mail server on the Rackspace Cloud.    It will be interesting to see how this performs vs. our dedicated server.  It feels much speedier already, but we’ll see how that goes as time passes.  Stay tuned.

P.S. Much thanks to Santosh Rao w/ Zimbra Support for his help during this process.

Update: Whoops!  One snag.  Road Runner is blocking the IP address of this new server – so we can’t get e-mail to road runner accounts!  http://security.rr.com/cgi-bin/block-lookup?67.23.29.240 Working w/ RackSpace to resolve this, though it is cumbersome via a chat-only support system.  Wish it had a “ticket” system like classic Rackspace.

Today I began migrating our Zimbra mail server from our Rackspace dedicated server to the Rackspace cloud.  This is going to be an ongoing post that I intend to update throughout the process.

Migrating from: RedHat EL 5 32 bit

To: Rackspace Cloud Ubuntu 8.04 LTS 64bit

This mail server only has about a dozen users.  Still, I’m going for the 2gb cloud option as I want nice performance.  We rely on e-mail more than any other form of communication at Tipit.  I am following the instructions here:

http://www.zimbrablog.com/blog/archives/2007/10/moving-zcs-to-another-server.html

The first ting that concerns me is going from 250gb to 80gb.  Sure enough after moving everything over, the drive is at 60% capacity.  May need to look at bumping up the storage / increasing the server size since Rackspace cloud doesn’t let you add more than the allotted storage to the various instances.

Everything went very smoothly and being in the same data center the file move was extremely fast.  The final step:

As Root: rerun the installer without the -s option

just blew up on me.

The system will be modified.  Continue? [N] Y

Shutting down zimbra mail

Backing up the ldap database…failed.

./util/utilfunc.sh: line 1221: /opt/zimbra/openldap/sbin/slapcat: No such file or directory

After some further testing, this file does appear to be there, though it’s a symbolic link:

ls -la /opt/zimbra/openldap/sbin
total 8
drwxr-xr-x  2 root root 4096 May  3 22:30 .
drwxr-xr-x 10 root root 4096 May  3 22:30 ..
lrwxrwxrwx  1 root root   16 May  3 22:30 slapacl -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slapadd -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slapauth -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slapcat -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slapdn -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slapindex -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slappasswd -> ../libexec/slapd
lrwxrwxrwx  1 root root   16 May  3 22:30 slaptest -> ../libexec/slapd

I checked out the utilfunc.sh script and tried to run it by hand:

/opt/zimbra/openldap/sbin/slapcat -f /opt/zimbra/conf/slapd.conf -b ” -l /opt/zimbra/openldap-data/ldap.bak
-su: /opt/zimbra/openldap/sbin/slapcat: No such file or directory

Same result.   Checking the dedicated RedHat original server, the structure looks identical.  Permissions look identical.  Symbolic link looks identical.  There, taking a backup of LDAP using the same command works fine.

Alse tried:

/opt/zimbra/openldap/libexec/slapd -f /opt/zimbra/conf/slapd.conf -b ” -l /opt/zimbra/openldap-data/ldap.bak
-su: /opt/zimbra/openldap/libexec/slapd: No such file or directory

Same result

Opened a support ticket with Zimbra.  Rolled back and put original zimbra mail server back online.

Just received a very speedy response from Zimbra support:

Ideally you should have followed : http://wiki.zimbra.com/index.php?title=Network_Edition:_Moving_from_32-bit_to_64-bit_Server

I am not sure if what your are following is a very good idea if migrating between architectures. Things are bound to fail because the underlying binaries are pre compiled 32 bit binaries. It would be best if you remove the current install and do a clean install of the same version as of the old server and copy over the store,index,ldap and mysql data manually. The steps are outlined in the above wiki.

Do let me know if you run into issues.

Regards,

Santosh Rao
Zimbra Network Support

Ok, fair enough, though the symptoms of the symbolic links not working in Ubuntu don’t seem to be a 32 bit vs 64 bit binary issue.  But, sure, let’s follow the other method and see what happens.