To get rid of dynamicproxy (which is currently a SPOF) before the grid is fully shut down, we could route traffic to tools running on the grid via the kubernetes ingress. We can have webservice-runner create a manual Service object with its endpoint configured and and Ingress object to route the traffic to it like we do for tools running in Kubernetes. After that we should be able to remove dynamicproxy by either removing the lua routing from nginx or by moving TLS termination to the haproxy nodes and skipping the proxy nodes entirely.
Description
Details
Subject | Repo | Branch | Lines +/- | |
---|---|---|---|---|
Route Grid engine web requests via Kubernetes | operations/software/tools-webservice | master | +234 -75 |
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | None | T90534 Make toolforge reliable enough (tracking) | |||
Open | None | T283948 Eliminate single point of failure from Toolforge front proxy | |||
Resolved | taavi | T282975 Create Kubernetes ingress for tools running on the grid engine to remove dynamicproxy |
Event Timeline
Change 697096 had a related patch set uploaded (by Majavah; author: Majavah):
[operations/software/tools-webservice@master] Route Grid engine web requests via Kubernetes
Mentioned in SAL (#wikimedia-cloud) [2021-06-14T21:29:19Z] <bstorm> deploy package with the staged patch to switch away from os.execv to QA in toolsbeta as toollabs-webservice version 0.75 T282975
Mentioned in SAL (#wikimedia-cloud) [2021-06-14T22:21:28Z] <bstorm> push docker-registry.tools.wmflabs.org/toolforge-python37-sssd-web:testing to test staged os.execv (and other patches) using toolsbeta toollabs-webservice version 0.75 T282975
Change 697096 merged by jenkins-bot:
[operations/software/tools-webservice@master] Route Grid engine web requests via Kubernetes
Mentioned in SAL (#wikimedia-cloud) [2021-10-20T12:51:18Z] <majavah> rolling out toolforge-webservice 0.78 T292706 T282975 T276626