Atlas
- A brief description of what you are looking to do
- How you think this will help
- Why this matters to you
-
Autoscaling improvements
Hi, we have tried BETA of Autoscaling feature and we have some thoughts how to make it better. In its current setup its not really suitable for our production workloads. Here are some thoughts how to make it better:
- Define separate scaling steps
At them moment the scaling step is always 1. Going from M10 -> M20 which is not really suitable for burst loads where going one step up might not be enough. Same goes for rapid scaling down
Example:
Scale range = M10 - M50
Scale step up 4 = (M10 -> M50)
Scale step down 2 =…61 votes - Define separate scaling steps
-
Allow auto-scale for IOPS.
Allow auto-scale for IOPS.
10 votes -
Autoscaling - CPU/ Memory Utilization duration monitor - Minutes
Hi, Currently there is no option to initiate auto scaling during heavy workload/ requests comes and the cpu/ memory utilization is 75% for the even within
- 5 minutes
- 15 minutes
- 30 minutes
- 45 minutesThe system has to wait for 1 hour to autoscale to next tier.This one of the concerns for the heavy workload applications.
Idea:
Add a custom duration option as
- minutes also while monitoring the cpu / memory utilization
Example : When a CPU / Memory utilization is above 75% for the past 15 minutes , move the cluster to the next tier…
7 votes -
Add "scheduled scaling/suspend" option for certain clusters
There should be an option to schedule scale up/down or suspending/resuming at specified time & day.
For example, DEV environments could be auto-suspended at 6pm and resume at 8am on weekdays, and be suspended all weekend.
Production clusters could be scaled down e.g. from M80->M60 every evening and on weekends rather than being suspended completely.
This would help save money but would need to be coordinated with the existing auto-scaling based on load.
This would need to be at the cluster level since sometimes different clusters in a single project could have different workload patterns.
7 votes -
Time based auto scaling
Our production server workload is less after 5 pm till the next day morning at 6 am CST times. We are running MongoDB in full capacity even on off-peak times. It's not possible for someone to manually downscale the cluster during off-peak and I don't think the CPU utilization based auto-scaling is working for us. I would like to choose auto-scaling between CPU based or time based as set my own time window and choose the auto scale/down-sclae.
6 votes -
Add " Auto-scaling" events of the clusters to the Activity feed
Auto scaling is a great feature and as of now if any cluster gets scaled up/down an alert email will be sent to the email list provided in the Access Manager . However , in the activity feed we are unable to see them .
Use case:
For instance if an user want to know how many times the Auto-scaling happened and at what specified time that event occurred on a particular cluster there is no definitive answer as the only way is to check the email alerts that's been received.
It would be great if there is a way…
5 votes -
make auto scaling alert message granular
When a cluster is changed due to auto scaling a message is sent to all project members. Project member cannot unsubscribe to this.
Request :
Provide the option, like all other alerts, to change the the alert settings for auto scaling4 votes -
Allow more aggressive downscaling of clusters
Autoscaling clusters down a tier doesn't seem to work very well. Even under rather low memory usage, it doesn't seem like the procedure really ever triggers.
3 votes -
Scaling cluster connections independent of cluster size/tier
Currently whenever you want to increase Atlas cluster connections, you have to scale up Cluster Tiers. When an application doesn't require additional disk storage but only additional connections the additional cost is difficult to justify.
1 vote -
Profiles for Autoscale heuristics
Case #00731969
Is it possible to tweak the scale up settings so that my clusters scale up faster than they do today?
I think it is best to analyze my clusters and figure out how long they are staying above 70% and then average it out, and if equal to or greater than that average, I would like my clusters to scale up.
Once scaled up, I want it to remain scaled up for the next ~2 hours at least and only scale down if usage gets really small enough for it to justify scale down.
Right now, the auto…
1 vote -
connect with slack
Whenever the database is about to scale, give an alert via slack to the team.It will be instant and currently, alerts are over email if you could provide it over slack the tech team responsible for particular will be notified.
1 vote -
Unevenly Scale Global Shards
Globally distributed shards should be able to scale unevenly. We have uneven loads across the world and sometimes spikes can hit a single region. Autoscaling for that single region scales everything and increases costs unnecessarily
1 vote -
we should be able to develop library of ML algorithms as plug-ins
we should be able to plug in the already developed algorithms say from GitHub or any other repository
1 vote
- Don't see your idea?