Certain jobs are related to videoscalers and must run on videoscalers hosts exclusively, so we need an LVS endpoint configured for them in order to submit the jobs from ChangeProp
|operations/puppet : production||lvs::configuration: Add configuration for the videoscaler service|
|operations/dns : master||Add entries for the videoscaler VIP|
|operations/puppet : production||conftool-data: Add missing data for videoscalers|
|operations/puppet : production||role::mediawiki::videoscaler: add TLS termination|
|operations/puppet : production||jobrunner_tls: generalize to support videoscalers as well|
|Resolved||Pchelolo||T157088 [EPIC] Develop a JobQueue backend based on EventBus|
|Resolved||Pchelolo||T190327 FY17/18 Q4 Program 8 Services Goal: Complete the JobQueue transition to EventBus|
|Resolved||Pchelolo||T188947 Create an LVS endpoint for jobrunners on videoscalers|
I'd take the chance we have to do this to do as follows:
- Add mediawiki::multimedia to the jobrunners
- Add a second LVS IP, to be served from the same cluster, to use for videoscaling. This will guarantee we evenly distribute videoscaling jobs across all the backends
- Fold the old videoscalers in the new cluster, for added capacity
the differences between the two clusters are small enough (basically, some apache config we want to remove / unify anyways) that we really don't want to keep up with the cost of separation between the two.
Add a second LVS IP, to be served from the same cluster, to use for videoscaling. This will guarantee we evenly distribute videoscaling jobs across all the backends
Do we even need the second LVS in case it will be the same exact cluster?
Yes, so that scaling and transcoding jobs are balanced across the cluster between themselves rather than being in the same pot as with the other (comparatively insignificantly CPU-intense) jobs.