Mercury 0.8-Beta released

Events happening in the community are now at Drupal community events on www.drupal.org.
Greg Coit's picture

We are please to announce the 0.8-Beta release of Mercury AMIs for AWS. Included with this release:

  • Removed the predefined admin user from drupal. Now the first login defines the admin user
  • Moved drupal cron from the root crontab to /etc/cron.d/drupal for more flexibilty
  • Fixed as postifix misconfiguration bug
  • Now using symlinks to store mysql, varnish (and now) tomcat data to /mnt rather than moving the actual directories
  • A new script to configure varnish, php, tomcat and apc memory usage based on system RAM

If you choose to create your own server, the install instructions at http://groups.drupal.org/node/25425/ have been updated with the above changes.

As always, please post suggestions here and bugs to https://launchpad.net/projectmercury.

The AMI IDs are:

US 32-bit: ami-3af61453
US 64-bit: ami-2cf61445
EU 32-bit: ami-8f0922fb
EU 64-bit: ami-830922f7

Comments

Thanks!

toemaz's picture

Damn, I just upgraded yesterday to 0.71 ;-) It went smooth, not more than one minute downtime since I had to switch the EBS to another zone. Otherwise it would have remained up and running.

Anyway, thanks for the great work!

Sweet!

StuddMan's picture

I was just about to build 3 new EC2 servers for a site going live next week and was going to use 0.71. We are deploying our first simple cluster on Project Mercury with 2 Web-servers and a single DB server with EBS. We already have one major site running a split config with 1 web 1 db. Expecting 80K-100K users on launch day.

Someone can correct me if I'm

dalin's picture

Someone can correct me if I'm wrong, but Mercury will not work as a clustered web head without some modifications. Mercury is using APC for object caching which can only work with the local PHP. You'll need to switch to Memcache residing somewhere accessible to both servers. That probably shouldn't be one of the web heads, unless you split bins equally between them.

--


Dave Hansen-Lange
Director of Technical Strategy, Advomatic.com
Pronouns: he/him/his

My understanding is that you

StuddMan's picture

My understanding is that you would use memcache if you were clustering MySQL. The web servers can run autonomously with their own instance of Drupal and use APC for local PHP optimization. We will use the elastic load balancer to manage the load between the 2 servers. Will be setting it up today to test.

Mercury 0.8-Beta released not sending mail

StuddMan's picture

Deployed 3 instances of Mercury 0.8-Beta released and mail isn't working, Any ideas where I should start looking?

How about checking the spam folder

StuddMan's picture

Well it turns out the servers are sending mail fine but everything is going to spam now. Not sure why we didn't have issues with our .0.7 beta boxes like this. Going to start digging into EC2 outbound mail soultions to avoid being flagged as spam.

My domain & server tweaks for anti spam

toemaz's picture

List of checkpoints for mydomain.com, which uses Gmail as mail provider:

  • SPF & Sender ID:

    mydomain.com TXT "spf2.0/pra a include:aspmx.googlemail.com -all"
    mydomain.com TXT "v=spf1 a include:aspmx.googlemail.com -all"

  • Return-Path: by default, mail sent from a server uses the return path from the server which is in case of the EC2 server: www-data@ec2-xx-xxx-xxx-xxx.compute-1.amazonaws.com
    Luckily, Drupal has a module for that http://drupal.org/project/returnpath
  • /etc/hosts: if there is only one domain residing on the server!, add the line
    [YOUR EC2 IP ADDRESS] mydomain.com
  • Postfix: if the server is only spitting mails from one domain, edit /etc/postfix/main.cf
    myhostname = mydomain.com
    And restart postfix
    $ /etc/init.d/postfix restart
  • Fill in this form http://aws.amazon.com/contact-us/ec2-email-limit-request/

For the future hopefully:

rDNS: ask ISP of the server (Amazon in this case) to set the rDNS/PTR of the server. Again, only if there is only one domain on that server. Amazon doesn't have a solution for this yet, so fingers crossed since this is the most important setting to avoid that your mail is recognized as spam.

Feel free to make a wiki page of this list in this group.

Update regarding rDNS/PTR on ec2

toemaz's picture

It's now possible to request the rDNS/PTR record for your ec2 server by filling in the form via Increase your Amazon EC2 limits > Emails

You might want to go with the

brianmercer's picture

You might want to go with the less restrictive ~all instead of -all so that mail that is not coming from the A record or from google is treated more suspiciously, but not necessarily blocked. http://en.wikipedia.org/wiki/Sender_Policy_Framework

-all vs ~all

toemaz's picture

I always use the gmail smtp to send mails. Unless your outgoing mail is higher than 500 per day, the gmail smtp is a good option to use. So -all is good and avoids someone is spoofing your domain.

High performance

Group notifications

This group offers an RSS feed. Or subscribe to these personalized, sitewide feeds: