What are the impacts of bluemix auto-scaling in terms of resource management. For example if a runtime is specified with 1 GB of memory and auto-scaling is set to 2 instances, does the application consume 2 GB? Same question for the disk allocated for the runtime? Are logs from the various instances combined automatically? If an instance is currently serving a REST request (short), how does Auto-Scaling make sure that the request is not interrupted while being served?
Asked
Active
Viewed 167 times
1 Answers
3
When you say, "a runtime is specified with 1 GB of memory and auto-scaling is set to 2 instances" I assume that you set your group/application up such that each instance is given 1 GB of memory and you are asking what will happen if the Auto-Scaling service scales up your group/application to 2 instances.
Memory/Disk
For example if a runtime is specified with 1 GB of memory and auto-scaling is set to 2 instances, does the application consume 2 GB? Same question for the disk allocated for the runtime?
Yes, your application will now consume 2 GB of your total memory quota. The same applies for disk allocation.
The Auto-Scaling service will deploy a new instance with the same configuration as your existing instances. If you've set up your group/application such that each instance gets 1 GB of memory, then when Auto-Scaling increases your group's instance count from 1 to 2 your application will now consume 2 GB of memory, assuming that adding another GB doesn't go beyond your memory quota. The same applies with disk allocation and quota.
Logs
Are logs from the various instances combined automatically?
Yes, the logs are combined automatically.
Cloud Foundry applications combine logs as well. For more information about viewing these logs check out the documentation.
The IBM Containers service sends logs to IBM's Logmet service. For more information check out the documentation.
Handling REST requests without interruption
If an instance is currently serving a REST request (short), how does Auto-Scaling make sure that the request is not interrupted while being served?
Adding an instance to the group/application: no interruption
If an instance is being added to the group then there will be no interruption to existing requests because any previously existing instances are not touched or altered by the Auto-Scaling service.Removing an instance from the group/application: possible interruption
At this time, the Auto-Scaling service does not support protecting ongoing requests from being dropped during a scale down operation. If the request is being processed by the instance that is being removed, then that request will be dropped. It is up to the application to handle such cases. One option is your application could store session data in external storage to allow the user to retry the request.
Additional Information
There are currently two different Auto-Scaling services in Bluemix:
- Auto-Scaling for Cloud Foundry applications exists in all Bluemix regions and is available as a service you bind to your existing Cloud Foundry application.
- Auto-Scaling for Container Groups currently is available as a beta service for the London region in the new Bluemix console.
The answers to your questions above are applicable to both services.
I hope this helps! Happy scaling!
John Starich
- 619
- 11
- 15
Janek Lehr
- 31
- 1
- 4
-
Hi. Thank you for the quick answers! Do you know if there is a plan to resolve the possible interruptions when an instance is removed? For example, a potential strategy would be to starve the instance that is targeted for a shutdown and avoid sending requests to that instance for a specified period of time. That would avoid problems when requests are of short duration anyway... – D Roy Jun 22 '16 at 17:27
-
Eliminating possible interruptions when an instance is removed is one of the areas we are looking to improve. We will continue making improvements to the performance and reliability of the service, as well as rolling out new features like this one. – Janek Lehr Jun 24 '16 at 14:48