Blue-Green Release Rollout and Auto Scaling

At Apigee, we are continuously improving product reliability, availability, and scalability for our customers. As part of those efforts, we are updating the way we do our software rollouts. Our goal, as with these kind of changes, is alway to have no impact to your API runtime traffic and no action required on your part.

We are doing the updates to help customers realize the following benefits:

  1. Increased improvements in reliability for your runtime traffic during releases.
  2. Better and automatic handling of gradually increasing API traffic without any pre-planning or provisioning.*
  3. Traffic origination from fewer IP addresses to your backend target servers, resulting in easier management of Apigee whitelisted IP addresses.

What’s new?

  1. Blue-green release rollouts: Blue-green release rollouts will enable us to gradually roll out new releases of our software by having the current release and new release simultaneously active with traffic gradually being migrated over from the current release to the new release. Migration of traffic will be based on analysis of the behavior of both releases to ensure there is no regression detected in the behavior of the new release.
  2. Network Address Translation (NAT): Given many of you rely on IP Whitelisting for increased security, we are introducing a NAT layer between our message processors and your backend target servers. This allows the message processors to spawn dynamically and to continue to reach your backends without any disruptions while maintaining the security afforded by IP whitelisting. We are reusing existing message processor IP addresses, which requires no changes on your part.
  3. Auto scaling: Thanks to the NAT implementation, along with additional monitoring and provisioning capabilities within our infrastructure, we are also able to auto scale our API runtime layer (message processors). This ensures we have sufficient capacity in place to handle your growing API traffic needs.

What changes will you notice?

  1. Number of message processors** : Both blue-green release rollouts and auto scaling features rely on elastic infrastructure and will result in instances of our message processors dynamically appearing and disappearing. If you have any custom monitoring alerts, please update them accordingly.
  2. Traffic to your backends from fewer IP addresses: With the NAT layer, we need far fewer IP addresses than the number of message processors active at a given time. This means you will see traffic originate to your back-ends from fewer IP addresses than you would have seen previously.

This update will occur until the end of July, during the currently stated release windows. More information on the release windows and for future notifications like these, please visit and subscribe to status.apigee.com.

As usual, if you notice any issues with or changes to your API traffic during these change, we request you to contact our support team immediately.

-------------

*Auto scaling is only available for customers on a call volume-based licensing plan and not available for customers on a capacity-based licensing plan. Please contact your Apigee sales rep or sales@apigee.com if you wish to discuss any changes to your current licensing plan for Apigee products.

** For customers with a capacity-based licensing plan, the provisioning will be done based on the licenses purchased.

Comments
Not applicable

Hi Floyd

What about Apigee Edge in Private Cloud ? Does it support features mentioned above auto-scaling and blue-green deployment ?

Thanks,

Reza

jonesfloyd
Staff

Hey @Reza Motevallizadeh - This applies only to cloud right now. I'm not sure about plans or timelines for Private Cloud. It may be worth you asking that as a separate question to let others weigh in.

Not applicable

Hi Floyd

I have posted my question and I appreciate if you can let me know if you had some idea around Mr Cassandra and Zookeeper?

I am having difficulty to manage these guys in our environment. Specially with our complex CI/CD tool.

Thanks,

Reza

jonesfloyd
Staff

Hi @Reza Motevallizadeh - Thanks for posting the separate question. It would be worth updating that post to specifically call out C & Z. You're likely to get a more focused answer that way. Apologies that I don't have a definitive answer for you. I'll have to defer to the answers that come in on the other post. Thanks.

akoo
Participant V

Hi @Floyd Jones, are there more details on autoscale? E.g., when does it trigger for more resources? When does autoscale trigger for less resources? Are there any conditions for which autoscale wouldn't trigger for a volume-based licensed customer?

Not applicable

Hi @Floyd Jones, Could you provide details about custom monitoring alerts. I understand, Customer can't monitor MP instances in public cloud.

Not applicable

My understanding is autoscaling happens based on average CPU usage across MPs. Current number of requests is not a criteria to consider it. @Floyd Jones pl confirm.

Version history
Last update:
‎06-23-2016 02:46 PM
Updated by: