Hi Matthew, In answer to your questions (from Airbnb's perspective):
1. Regarding a production environment it's probably highly dependent on your use case. At a minimum I would suggest having at least 3 web-servers (we use Gunicorn <https://gunicorn.org/>) and 3 Celery workers for redundancy. We use Datadog for monitoring/alerting which has helped us track load and scale accordingly. We use Kubernetes for deployment which includes some elasticity, though currently we're running 3 web-servers and 5+ Celery workers. The later may be overkill though we're trying to debug an ongoing QueuePool issue. 2. We use AWS to house Redis, Superset, Celery, etc. which all run from the same region <https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-regions-availability-zones.html> . 3. Our servers run on 4 cores with 10 GB of memory. Note we haven't optimized this configuration. 4. We have low thousands of weekly active users for Superset and SQL Lab. -John On Mon, May 20, 2019 at 12:24 AM Matthew Mutee <mute...@gmail.com> wrote: > Hi dev, > > We have been testing superset for the last couple of months and now want to > set it up in production. Could you please share what sort of resourcing > considerations you have made for your superset installation. > > I am interested in: > > 1. Number of servers used for load balancing. > 2. Placement of applications; redis, superset, celelery, nginx etc. > 3. Actual resources in the servers; memory, cores etc > 4. The average users accessing superset per day. > > Thanks :-) > > Regards, > Matthew M. >
