Right now we have the following blockers that would preclude from migrating low-traffic services to Katran:
* Direct routing mode being used to forward traffic to realservers
* Weighted round robin (wrr) scheduler
IPIP encapsulation can be adopted by non-k8s based services behind the low-traffic load balancer in the same way as it was done for high-traffic2 services on T367312.
Moving away from wrr is feasible using the mh-port flag as mentioned on T368545, a small patch is required on pybal to support the mh-port flag to use `-b mh-port` flag on ipvsadm.
This is the list of services that could benefit from this migration (k8s based services are excluded from this migration):
| task | service | contact|
| ----- | ----- | ----- |
|| apus| #data-persistence|
|| swift |#data-persistence|
|| swift-https| #data-persistence|
|| thanos-query| #data-persistence|
|| thanos-web |#data-persistence|
|| thanos-swift |#data-persistence|
|| datahubsearch| #data-platform|
|| druid-public-broker | #data-platform|
||schema|#data-platform|
|| docker-registry| #serviceops|
|| jobrunner | #serviceops|
|| videoscaler | #serviceops
|| kartotherian| Core Platform|
|| kartotherian-ssl | Core Platform|
|| restbase-backend| Core Platform|
|| restbase-https|Core Platform|
|| kibana7 | #observability|
|| prometheus| #observability|
|| prometheus-https |#observability|
||logs-api |#observability|
|| labweb-ssl | #cloud-services |
|| search | Search Platform / #data-platform-sre|
|| search-https | Search Platform / #data-platform-sre|
|| search-omega-https | Search Platform / #data-platform-sre|
|| search-psi-https| Search Platform / #data-platform-sre|
|| wcqs |Search Platform / #data-platform-sre|
|| wdqs |#wikidata-query-service|
|| wdqs-heavy-queries|#wikidata-query-service|
|| wdqs-internal|#wikidata-query-service|
|| wdqs-ssl |#wikidata-query-service|
This has been automatically generated using the following prompt:
```lang=bash
sybil:puppet vgutierrez$ yq '.catalog
| to_entries
| map(select(.value.lvs.class != "high-traffic1" and .value.lvs.class != "high-traffic2"))
| map(select(.value.lvs.conftool.cluster != "kubernetes"
and .value.lvs.conftool.cluster != "k8s"
and .value.lvs.conftool.cluster != "aux-k8s"
and .value.lvs.conftool.cluster != "dse-k8s"
and .value.lvs.conftool.cluster != "kubernetes-staging"))
| map(select(.value.lvs.conftool.service != "kubesvc"
and .value.lvs.conftool.service != "kubemaster"))
| map(select(.value.lvs.conftool.cluster != null))
| map(.key)' hieradata/common/service.yaml | sed s/-/[]/
```
contact information obtained (partially) with this oneliner:
```bash
vgutierrez@carrot:~/wikimedia.org/operations/puppet/hieradata$ yq '.catalog
| to_entries
| map(select(.value.lvs.class != "high-traffic1" and .value.lvs.class != "high-traffic2"))
| map(select(.value.lvs.conftool.cluster != "kubernetes"
and .value.lvs.conftool.cluster != "k8s"
and .value.lvs.conftool.cluster != "aux-k8s"
and .value.lvs.conftool.cluster != "dse-k8s"
and .value.lvs.conftool.cluster != "kubernetes-staging"))
| map(select(.value.lvs.conftool.service != "kubesvc"
and .value.lvs.conftool.service != "kubemaster"))
| map(select(.value.lvs.conftool.cluster != null))
| map(.key)' common/service.yaml |egrep -v "^\[$|^\]$" | sed s/\"//g | sed s/,// | sed "s/ //" | xargs -I % sh -c 'git grep role_contacts | grep % || echo "contacts for % not found"'
```