On the beta cluster our deployment servers are deployment-tin and mira and have:
As part of T144578: Migrate deployment servers (tin/mira) to jessie we created a new host deployment-mira02 and played it nice with the labs resource with an m1.medium:
That does not work though. We run scap on it for mediawiki which generates l10n files and need bunch of CPU to rebuild in parallel. That is spike usage and not going to use 100% of real CPUs. 8 vCPU seems appropriate.
2GB for RAM and 6G for the disk cache would be good enough for a total of 8G
For disk, 20G is the system partition we need a /srv that is larger than 20G. So a total of 60G looks about right.
Hence I am requesting the creation of a new flavor for the deployment-prep labs project:
I suggest the name 8c8m60d which I have seen used somewhere.