Group of jobs running on glideins prepared for different jobs
2 frontend groups G1 and G2 both match and request glideins from the same entry E.
G1 has proxy P1 which gives certain privileges including 4 days lifespan
G2 has proxy P2 which gives only 1 day lifespan.
Jobs from G1 ended up running on E with glideins using proxy P2 and crashing prematurely
This has been reported via email and INC000000830732
This seem to happen because the factory considered equivalent P1 and P2 (2 interchangeable proxies for the same resource) and used only P2 for a certain amount of time.
This could happen with jobs (different autoclusters) from the same group. But should not happen with different groups.
There is a match expression and a start expression in the group configuration.
The match expression is used to request jobs and the start one to match glideins.
An attribute could be used to restrict the start expression to match glideins only with the specific proxy
I'm wondering if this is due to [#11142] and an overlap at the frontend or if the overlap of the credentials happened somewhere else.
Asking for clarification about the configuration.